The conclusion is simple : when all lenses are stopped down to the same numerical aperture, namely we used f/11, the tolerance on film positioning, measured by reading the sharpness of test targets, is the same for all focal lengths stopped down at f//11, exactly as predicted by the geometrical model and the approx tolerance we derived from our tests matches perfectly the classical value of plus or minus N.c, no mystery.
Emmanuel, nice tests. Hard to argue with practical results. You now have me wondering if I should purchase a few dedicated digital lenses for the 6x9 back.