Perfect, thanks for sending the files. It's clear what happened in this particular case. I am seeing one stuck note G4 in this file. As you can see in the screenshot, the computer did not receive a MIDI note off (actually a note on with velocity 0) for the first G4.
I suppose Band-in-a-Box should automatically be setting the maximum duration of the note if it doesn't receive a note off before the same note is played again?
I would be interested in seeing a .SEQ (RealBand) with a MIDI recording to see how RealBand is handling this. It would also be good to look at another sample recording to see if there are any patterns.