Borrowed my friend's laptop/interface for a test.
I generated five white noise files: at 0dBFS, -0.5dbFS, -1dBFS, -3dBFS, and -6dBFS. I also chose two music files: one mp3 ("
Mexico City" by Jolie Holland), and one wav ("
The Dream that You Are" by Libby Kirkpatrick), all 44.1k. Noise files were mono (well, stereo with identical track on both sides, due to accident.)
Interface used was a
Universal Audio Apollo Twin. Did an analog loopback via balanced 1/4".
Played from Reaper, recorded in reaper. Played from iTunes, recorded in reaper.
I had to adjust the input gain slightly on the interface when playing through iTunes, but it may have had more to do with the coarse-grained GUI for adjusting the input gain than anything else (I had to set it at the last "click" before clipping when playing the 0dBFS file... i.e. it might have been a tiny rounding issue or something and so I had to go to the next-lowest click for iTunes... not sure.). This resulted in a difference of 0.7dB between the reaper and iTunes results, but this 0.7dB relative difference was preserved exactly across all test results.
So, no real difference in levels. There was no difference in the stereo field, judging by "lissajous" vectorscope. There was no difference in long-term average EQ. There was no audible difference at all.
Short version: I could find no difference in the results at all.
So, with my friend's mac laptop and that interface, I found nothing to imply that iTunes was messing with the sound in any way.
I think the OP needs to do the blind testing and get back to us. :-)