Apart from the intellectual interest of trying to understand how all the elements of the imaging chain affect the final outcome, the important message perhaps is that "Where's the diffraction limit?" is not as useful a question as "Where's the sharpest aperture?", whether the falloff in sharpness at smaller apertures is due to diffraction or to other effects or a combination. The empirical data amassed by Perez and others certainly suggest that most MF lenses, including the Tessars and Xenars used on Rolleis, peak in resolution around f/8-f/11. But the drop in resolution at f/16 isn't very dramatic, and a lot of practical user experience suggests that the actual visible effects are pretty minimal up to f/22 if not beyond.
Thanks for the pointers to Perez's article and others.