Reading between the lines of this non-statement statement you don't list your secondary or tertiary goals, or give any technical information.
So the average reasonably technically savvy reader of this long and winding thread might instead conclude that what is really going on is that rights holders are adding a bit of bass EQ and dynamic range compression to make tracks on iTunes "hotter" since they know most listeners are on earbuds or low quality speakers. Same thing has been done on radio for years. So MFiT means you are trying to dial that back so that the AAC encoder doesn't clip.
We can conclude this because hundreds of people on this forum have shown that any garden variety AAC256 encoder is capable of generating a file that is nigh on impossible to distinguish from the original CD in a blind ABX listening test. So really MFiT is not needed for the reason you describe, unless the "original source" is not actually the same as the "original CD".
You could easily refute my supposition in one clear sentence, but you won't....
I don't know who this mr. Stewart is, but there seems to be some confusion as to what 'Mastered for iTunes' actually is.
It is a purely technical requirement set by Apple, and is made out of two parts.
One: the file sent to Apple should be of the highest resolution possible. If you have a 24/96 master, send that, and don't down-sample it.
Two: That file shall have its levels set so that it won't clip as it is converted to AAC. This includes inter sample clipping.
That is all. Really simple, I can't understand that this hasn't been done all along.
They've also made a little
software package to help in the process, and it works very well. The RoundTripAAC AU plug-in even has a simple ABX tester included.
As you can see from the screen grab, it is clipping rather badly, even though the source file peaks at -0.3dBFS. I eventually had to attenuate it 3dB to make it come through clean. I also converted that lossless (CD) file to AAC 256 cvbr, with no attenuation, using iTunes, and the number of clipped samples matched what the plug-in predicted.
This phenomenon of lossy compression leading to clipping isn't a new thing, and isn't restricted to AAC. When I converted that same file to LAME V1, it ended up even worse. Lastly I checked the iTunes Store version of the same song (The Trip by Still Corners, in case someone wondered), and lo and behold, no clipping! Seems like the system works.
(other than this minor difference in amplitude, and of course some (inaudible) artefacts, the two commercial releases seems to be identical, as they should be)