Against the idea of the capabilities of instruments affecting what happens to tunes: in many traditions, vocal music is considered primary, and instruments simply approximate what the voice can do. So anybody playing the traditional corpus on whatever instrument will always have the vocal model in mind. This is true of Indian classical music, Arabic and Turkish art music, and, if Allan MacDonald is to be believed, piobaireachd. (His thesis is beyond me, but I have reason to believe he knows what he's talking about).

The more familiar a tune, the more likely that is to happen. People playing Scottish singalong tunes on the accordion are not going to leave out notes beyond the pipe scale just because they once heard a piper playing those tunes without them - they'll go back to the way they're sung. And people singing along with a piper will just ignore the funny intonation and octave shifts a piper might have to do; they know how the song goes. (Who would ever sing "Scotland the Brave" the way it goes on the pipes?).

In critical theory, this is the "emic"/"etic" distinction. Comparing sound files with frequency analyzers, or comparing the notated forms of tunes, is research at the etic level; figuring out why people perform them that way is the emic bit.