I got a good result in P37. I used 144:45:00 as the TIG and then loaded 0 for both R2 and R3 of N60. Then just after the PRO I cheated and used 0.1x time acceleration, just making sure that input speed wasn't a problem. And I entered this exactly, right after the PRO on N60: V24E N1E 3651E 70000E E. This sequence of inputs (same as in the CMP Checklist) uses as little inputs as possible. N1 is being read as 01 and the 3651 also works fine and is the same as if you had input 03651. And very important, don't forget the second enter after 70000 as V24 enters two lines of data. An enter without inputting anything is the same as if you had entered 0, so 00000E.
I also just tried it at 1x time acceleration and it also works, so it's humanly possible to enter the erasable memory change quick enough. As the P37 calculation runs fairly quickly with a TIG so close to reentry I wasn't sure if that was even the case, but it seems possible.
In either case I got this N81 (DV):
-00030
+00000
+00001
What the CMC and is doing here is that it copies a double precision value from fixed memory (which you can't change) to erasable memory (which you can change), pretty quickly after the PRO on N60. Only later on in the calculations (but not that long after) does it actually use that value. So in that time period you are able to change the value in erasable which the CMC uses as a sort of speed limit for reentry. The octal value entered is 70000, 00000 or rather 70000, 77777 (00000 is +0, 77777 is -0, but it doesn't matter which one you input). That is a negative number for a limit on the major axis, so that's why the procedure is called -MA. Before we found the Apollo 8 CMP Checklist I had actually figured out this procedure myself, using the value the CMC had hardcoded for -MA from Apollo 10 on:
V24 N01E
3651E
75725E
75417E
Works the same, just takes more inputs and more time so the CMP checklist inputs are definitely better.