I did find some songs that had some c1 commands, followed by some nonsense, then I found some more c1 commands just after that led to the correct offsets.
Looking over the data again, it looks like the very first C1 commands in the Gamecube IPL (C1 80 00 00 2F C1 81 00 01 70 C1 82 00 02 81 C1 83 00 03 7D) show also have end of track bytes, but they're not FF. They end with an E5 byte. I can't seem to find E5 documented anywhere in any BMS notes online, so I'm wondering if this terminator byte is exclusive to this early version of BMS.
Could you try running a sequence with a single channel and putting FF and E5 at the end to see what happens?
I didn’t look at any notes regarding BMS files, I’m doing everything just from trial and error and comparing to what I know of other sequence formats . Feel free to share some of these notes.
There seems to be a script in some of the sequences to determine what “state” your game is in, and this effects what instruments will play. For example, a percussive track might only start playing when a certain goal/objective has been met. I wish I could edit this because I’m unable to get the sounds to play. I will eventually work it out from brute force . Maybe this week I’ll have a try