The Middle Shogi Manual was compiled by the late George Hodges
This website lists some new insights into these problems

All problems were fed to the HaChu engine. This was running in tsume mode with the 'Allow repeats' setting, which puts the obligation to deviate always on the checking side. The 'Promote on entry' setting was switched off, which means that HaChu plays according to the rule that a piece that deferred promotion can promote on later moves that touch the zone, but not on the turn immediately following their deferral (unless they capture). Note that this is different from what the Middle Shogi Manual states, where the condition is that the next move with that piece cannot be a promotion if it isn't a capture.
Below the cases are discussed where HaChu's analysis disagrees with solutions or comments in the MSM. The issues can consider minor typos, a somewhat faster line, but also complete refutation of a given solution, and spectacular new solutions.
B5  The given solution is poor defense; with 2... K5c 3. DHx4d+ Kx4d 4. FKx2b K5c 5. FK3c K6b 6. FK4d K6a 7.+DH2e G5b 8. FK4a K6b 9. +DHx5b K7c 10. +DH7d K8b 11. +DH8d K9b 12. FK9f K8a 13. +DH4a K9b 14. +DHx10a K9c 15. FK9d mate gote can hold out much longer.
B7  The given solution is correct, but there are several ways to finish the mate quicker: 15. +SM7c K5f 16. +SM4f Kx4g 17. +SM2g Kx4f 18. +VM6d K5f 19. +SM6g mate or 17. +SM4e6c K4b 18. +VM6d BT5c 19. +VMx5c K4c 20. +SM6a mate
B8  The mentioned variation, 2... Kx6c is actually beter defense, as the Lion mate takes one move longer.
B9  For this problem the exact promotion rule is important. When promotions are allowed on stepping out of the zone, (possibly after a delay of a turn) 2. FL10d+ is one move faster than the given solution: 2... S9c 3. +FLx9c BT8b 4. +FLx8b K6a 5. FK1f Rx1f 6. +FL5b Gx5b 7. +DK4a G5a 8. +L2e DKx2e 9. Ln5b Phx5b 10. DK4c mate
Compared to this the given solution inserts the extra moves 2. +DH10d FLx10d to lure gote's FL to 10d, so sente's FL can leave the zone with a capture. These moves are a pointless delay when promotion had also been allowed when leaving the zone on a noncapture. So it seems that at least the person providing the solution assumes the latter is not allowed. The +DH on 7g is still needed in any case (or 2. FL10d+ would be refuted by 2... FLx10d), so it gives no clue as to what rule the composer used.
B11  The 'correct' solution can be improved to a mate in 10 through 5. +DK7c K9b 6. +P8b K10a 7. +DK7cx7a K11b 8. +DK+12a K11c 9. +DH12d K11a 10. +DK1a12a mate 9. +DK11a K12c 10. +DK1a10a mate
B12  The solution is flawed, because 3... K6b, presented as a suboptimal defense, is actually a refutation. The given variation does not end in mate, as the checking +DH supposed to deliver it can be taken by the +R at 4d. The first move of the given solution already spoils the mate. It is essential to leave the BT on 4e, to prevent the +R from capturing on 4d.
The correct solution thus is: 1. +Ky4d Gx4d 2. FKx6d Kx6d 3. Lnx8f K5e 4. +DH3g +R4f 5. +DHx4f Kx4f 6. Ln6h K3f 7. Lnx5h4h K2e 8. Ln4f K1d 9. Lnx4e3f K1c 10. Ln1e K1b 11. Lnx2e K2a 12. Ln4b K1b 13. Lnx4a3b
The suboptimal defense 1... BTx4d then continues like the given solution, for mate in 12: 1... BTx4d 2. Ln6e Px6e 3. +DHx6e K6b 4. +DH7e +R6j 5. FKx6j B6h 6. FKx6h +FL6g 7. FKx6g +VM6f 8. +DHx8d K6a 9. +DH8c K6b 10. +DH8b K6c 11. FKx6f G6d 12 FKx6f mate, which is faster despite of a number of 'futile interpositions'. (While 3... K7c continues as the given solution, also for mate in 12.)
B13  After 7. +DH6c it is a mate in 12 rather than 17:
7... K3a (7... K5a 8. FK5c +DHx5c 9. +DHx5c K4a 10. +L4c mate) 8. FK3f K2a 9. +DH2c FLx2c 10. +Lx2c +DH2b 11. FK3b K1a 12. FKx2b mate
B15  The given line is flawed, as after 3... K5b there is no mate. But when we play 3. P6f= in stead of P6f+ it works!
At first glance having a Tokin should always be better than having the Pawn, as the Tokin's moves include the Pawn move. But there is one thing a Pawn can do that a Tokin cannot: stand between two Lions without forming a 'capture bridge' between them! So the move 4. Ln5d which saves the day after 3... K5b cannot be refuted by 4... Lnx6dx5d when 6d is a Pawn!
B17  In the given solution 6... K4c is poor defense. With 6... K5c in stead it is mate in 18:
7. +DH2f S4d 8. FKx4d K6c 9. +DH9f K7b 10. +DH9d K6c 11. +DH8e K7b 12. +Px7c DEx7c 13. FKx5e K8b 14. +DH11e K9b 15. FK7b DE8b 16. +DHx8b K10c 17. FK10e K11c 18 +DH10b mate
B20  The given 8... K10e is very poor defense. It is true that it is custom to not count futile interpositions in tsume problems, like DE4d here, but 8... S7d does not fall in this category, as it is protected, so that it will cost sente a HF to remove it, a HF which will be sorely missed in the mating. So playing it can delayed the mate to a mate in 19:
9. +DHx7d +SMx7d 10. +Phx7d K9c 11. +P10c Kx10c 12. +L10h K9c 13. +L10d Bx10d 14. Rx12c+ K10b 15. +Phx10d K9a 16. +R12a K9b 17. +R10a K8c 18. +R10c BT11c 19. +Rx11c mate
B22  This problem seems to be flawed. The published solution can be refuted by 2... +Lx5g. Trying to repair the solution by letting the Kylin move to 5e through the other path, 2. Ky3e opens the possibility for the defense 2... SMx3e. (After 2. K6d this would bring us back to the published line.) This line gets excedingly interesting:
1. +DH5e Kx5e 2. Ky3e SMx3e 3. +DH3g K4e 4. +DHx3e Kx3e 5. FK3g K4e 6. FK4f K3d 7. FK3f K4c 8. VMx2a+ K5c 9. FK3a K5b 10. +VM2e BT4c 11. FK3b K5b 12. FKx4c K6d 13. FKx4d K7c 14. FK5e G6d 15. FK3c G6c 16. +VM2h
All this looks designed to shephard the King into the 'death trap' formed by the two enemy Tokins and its own Pawns, the White Horse waiting on the outside to strike through +Lx9d, and finish off the job together with the +SM that is already 'looking in'. But there is no mate here!
We note that the MSM gives the problem in Forsyth notation with a Silver on 6b, while the diagram in the appendix has a Copper drawn there. But in either case there is no mate. This inspired an idea for 'repairing' the problem by replacing the generals, however. It turns out a forced mate exists when we replace the Gold at 6c by a Silver, although there is no plausible way to ascribe such a substitution to a copying error. This makes it impossible to use this general as a permanent shield by shuttling it between 6c and 6d, (as in the moves 15... G6d 16... G6c in the line above). The 'fixed' problem then becomes a mate in 20, concluding the line above with
13... K7e (13... K7c 14. +VM8h K8c 15. +VM8b K7b 16. +Lx9d Kx8b 17. FKx6b S7b 18. FKx7b mate) 14. FK7g K6d 15. +VM2h K5d 16. FK5e K4c 17. +VM2e K4b 18. FK4d K3a 19. FK3c K4a 20. +VM4c mate
B23  A Lion can do everything a Kylin can. Except... It is subject to rules for its capture! This problem highlights this, with 5. Ky5c=. Should we have promoted here, then we would not have been allowed to take gote's Lion with 6. FKx9a (counterstrike against Lion) after gote captured our +Ky with his Phoenix.
B26  The given 3... +VMx2d is poor defense, and after the better 3... FLx2d, 9... K4c is poor defense. In this variation, after 9... K3c the King can escape from that corner, and the mate takes far longer: 10. Lnx5e4e K3b 11. Lnx3d K2a 12. L1b+ Lx1b 13. +P3a Kx3a 14. Lnx1b K4b 15. Ln2b K5a 16. Ln3b K6b 17. Ln4a K7c 18. Lnx5b6a etc., for a mate in 27 in total.
B27  The given solution obviously assumes the Kylin is not allowed to promote on its first move. Otherwise this move (i.e. 5. Ky4d+) would have been an immediate mate. But in stead sente goes through a cumbersome manoeuvre to lure the King to 4b, so that the Kylin can step unpromoted with check to a square where it can promote through capture. And the problem is obviously designed that way: for the given solution every piece is essential, while promoting the Kylin on an inzone move would still be a mate in 5 if we deleted the Lion at 1c and the +B at 10f.
So we can conclude here that both the designer and the solver of this problem assumed inzone noncaptures could not promote, even for a piece that had not moved on the previous turn.
B29  The mate can be achieved one move faster by leaving out the first move: 1. FK12k +Px12k 2. +DH11j +Px11j 3. +FL10i +Px10i 4. B9h +Gx9h 5. +VM8g +RCx8g 6. +SM7f P6e 7. BTx4e K5c 8. +L1g +VMx1g 9. +VMx1g S4d 10. +VMx4d Kx4d 11. +DHx3d K5c 12. +DH3e K4c 13. BT3d+ K5d 14. +DH4d mate
B29  Gote could delay capturing the +P a few more moves as it pushes in on his King, to make a longer mate.
B31  Like in B27 the published solutions refrains from promoting the SM on an inzone noncapture move (5. SM7d+), but sacrifices a Lion first to lure an enemy piece to the square where it wants to promote. Otherwise 4. Lnd7 Gxd7 would just be a pointless delay. In fact the Lion could be completely deleted from the problem without consequence!
So also here there is strong evidence that both problem and solution author assumed inzone noncaptures could not promote. This could be just because the SM was 'tainted' by deferring on the previous inzone capture on the immediately preceding move, where it could have promoted.
B33  The published solution avoids promoting the FL on leaving the zone with a noncapture. Which could have provided a faster solution (1. FL4e+ +L2c 2. DK3d+ +Lx3d 3. +L1d Gx1d 4. +FLx3d mate). This solution would also work without the +DH at 8c, and so is obviously also not what the problem designer intended.
B39  The defense 1... K6a can be improved to make it a mate in 12 (making it main line):
2. +DK2e4c Gx4c 3. +DK5b Kx5b 4. +DH9f Ln6c {after wich the lioncapture rules forbid you to play your Lion to 6c} 5. +DHx6c K4a 6. Ln6c K3a/b 7. Lnx6b5b K2b 8. Lnx4c4b K1a 9. Lx1b+ Kx1b 10. Ln3d K1a 11. Lnx2c3c mate
The given solution can be sped up to mate in 8 through 5. +DH5b B4b 6. Lnx4d K3a 7. +DK3e DK3b 8. +DKx3b mate
B41  Better defense is
3.. Kx6a 4. DHx4a+ K6b 5. +DH4b K6a 6. +R9a G7a 7. +DH5a K7b 8. +Rx7a K8c 9. +R8a K7d 10. +Rx8d K7c 11. +R8c mate
B42  Faster is 7. +DH5a +DHx5a 8. +SMx5a K8b 9. R6b+ BT7b 10. +R7c mate
B45  The published solution is refuted by 4... +R8h, and this flaws the problem as given, as there are no alterative mating lines. The problem can be repaired, however, by putting an extra sente Pawn on 8j, which blocks the refuting +R interposition. Unfortunately this extra Pawn hinders sente's +SM in the later stages of the mate, so that it takes even longer:
11... K9k (11... K8j 12. +BT8h K7k 13. Lnx9i K6l 14. +SM11l K5k 15. Ln7i K4k 16. Ln5i mate) 12. Lnx9i K8l 13. Ln10j K7l 14. +BTx7j +Rx7j 15. Ln9l K6k 16. +SM10k K6j 17. Ln8l K6i 18. +SMx12i K6h 19. Lnx7j K6g 20. +SM8i K6f 21. Ln6h K7e 22. Lnx5g5f K8d 23. Ln6f K8c 24. +SM5f K9c 25. Ln7d K10d 26. +SMx8f K11e 27. Lnx9d K12f 28. Ln10d K12g 29. +SM7g DE9g 30. Ln10e K12h 31. +SM10j K11h 32. Ln9g K12i 33. Ln10i mate
B46  Note that the problem can be repaired by replacing the FK at 2i by a +FL at 1j, preserving the published solution.
B49  The published solution is refuted by 7... K7c, after which there is no tsume. 8. Ln5d K8d 9. Ln6e K9c 10. Ln7d K10b 11. +P11c (11. Ln9d K10a {and there is no safe Lion check}) Kx11c 12. Ln9c K12d {and there is no way to pick off L12a, which is fatal later}
B50  Unlike what the MSM states, after 1... FKx1h there is a mate in 18. So this is actually the main line of the solution:
1. SM1h FKx1h 2. FL8c= Cx8c 3. +DKx8c Kx8c 4. Ln7e K9b 5. Ln7d K10c 6. Ln8c K11c 7. Ln10a K10d 8. Lnx12b K9e 9. Ln10c K9f 10. Ln11e and Lion mate on the 1file follows.