And in fact some of his tests show that this is not the case. He routinely tests AF repeatability without AFMA adjusting cameras to lenses and although the data shows AFMA is off, repeatability is pretty much the same no matter where your AFMA is set.
One other thing he discovered is that new lenses provide AF feedback to the cameras and only the new cameras make use of it (5d3, 1DX, maybe 6D but that came out after his tests were completed). So your best AF accuracy AND repeatability are with using a new lens with a new body. Other combinations provide just the "standard" old-school AF.

Canon hasn't really boasted much about this but Roger was cracked open a few lenses and found the sensors. The tests show that only the newer cameras make use of it.
There were several parts to this series but here is one that describes the AF repeatability. I'll leave it up to the reader to find the previous and future posts on the subject:
http://www.lensrentals.com/blog/2012/08/autofocus-reality-part-3b-canon-camerasI would also respectfully submit that in all likelihood, a camera WILL AF differently and with less consistency with varying lighting conditions and the type of light used. I will even go as far as to submit that AF systems will vary depending on the subject target used. However I am pretty certain that AFMA doesn't affect repeatability.
Have any of you found the same to be true?
Thanks
I've not noticed it. But I haven't looked for it either. The new lenses on the new bodies were tested by Rodger at lens rentals. I don't think he mentioned anything like this, and it would seems an unusual effect from such a system.