The sorts of data which can now be collected are making even the two-point AFMA on the 1D-X seem needlessly primitive. With the right software one can get a whole heap of AFMA data in, what, maybe an hour per lens--body combination? Distilling this down to one pair of AFMA values (or worse, a single AFMA value) is then in a needless compromise, especially given it's only a software issue. I agree with John's observation that the correction is non-linear. My preference would be to be able to provide to the camera exactly the sort of grids being posted in this thread -- the AFMA given a focal length and distance-to-subject -- and have it extrapolate between points.

(The cynic in me wonders if the AFMA feature is deliberately crippled. At the risk of opening a can of worms, third-party lenses seem to need more correction, so a comprehensive AFMA mechanism would allow any deficiences to be compensated for. Or, even with Canon lenses, if they calibrate them after the warranty period, they pocket a nice fee.)