Did you have to have the incorrect 3 files to compare to the correct file in order to determine the correct file?
If the correct file could be determined without having to look at any other files for comparison, then that shows that there's a unique solution to separating out the jitter (noise) from the music (signal), which I would not think is the case.
I haven't tested any of this stuff, but I'm very interested in results towards this end. Think about all of the jittery online sources (e.g., youtube, my computer-based tivo, bowling alleys) that could have their sound much improved if we could partial out the jitter from the source. But again, I tend to think that jitter gets embedded into the signal, and there's not a unique solution for separating the two after they've been summed together.
But I could be wrong. I mean, our ears can separate summed signals very well (spatial location of different instruments), so it may be possible to 'spatially' separate jitter from the signal...
-Clayton