OK, I've been trying things out and it now makes much more sense...

I started over with the .SGU from another band member, thinking I was starting from scratch, but then reaklising that the instrument/player patches were as before. I tried it, I can't off-hand remember whether I had no melody or the song regenerated and destroyed it. I was able to get it back simply my undoing and redoing a key change (the SGU's in F but the MIDI was in Eb).

As I switched back to F, I had a pop-up asking if it was OK to erase the mel;ody and soloist tracks as they're audio and can't be transposed. I answered no.

Obviouslt that's a useful clue, so I removed the RealTrack from the melody instrument, set a MIDI patch and played the song.... Pretty darned close to what I expected... melody in first chorues, then solos with no melody. The only obvious oddity was no melody for the final chorus ... that's likely some other silly I've done.

I changed the melody patch to one from Sforzando, which sounded better and still played as I expected.

Then I repeated what I'd done some time back and changed the melody to a RealTrack instrument. The melody remained intact as I'd noted earlier. It seems very likely, though, that tyhis is the root of my operator error, as from here in, I think it's controlled more by the RealTrack than the MIDI ... I'm fine until I try to take this forwards, whereupon the rules for RealTracks trump the rules for MIDI (I think!).

Again I removed the RealTrack, played the now standard MIDI patch and then again applied the Sforzando patch. All behaving reasonably with the probable exception of that final chorus. I'll explore that further.

Jazz relative beginner, starting at a much older age than was helpful.
Kawai MP6, Korg M50, Ui24R, Saffire Pro 40.
AVL:MXE Linux; Windows 11; Win8.1: Scarletts
BIAB2022 UltraPAK, Reaper, a bunch of stuff.