Well, certainly, on the basis of what Jason has done, these files MUST have been MidiSoft, as they seem to have converted fine.
I've taken the liberty of having a listen to the .MID versions. D1 seems to be a partial version of a version of the same piece. D2 seems to be slightly different, slightly re the instrumentation, and slightly again re some of the variations, but D2 I'd say is complete as there's a start, a middle and an end. Musically, D1 seemed slightly nicer, but D2 is still fine and has the extra benefit of being complete.
Very nice work, really.
Given the respective size of the files (.SNG and .MID), yes, there is a LOT of data that has been lost, but clearly there is a LOT of data in the .SNG that has nothing to do with MIDI. No idea what it might be?
The midi files clearly included instrument settings, and I looked in the raw midi files and saw that, for example, 'Flute' and 'Bass' were there as text. So I did a search of the data in the .SNG file and found the same text, but well through the file, widely spaced out (Flute is at &H7528 and Bass is at &H11A04). Each is followed closely by data, some of which MIGHT be midi data. But it's FAR from obvious.
Keep up the great work!
Geoff