Thanks for good explanation. So calibration just fixes autofocus issues?
I'm actually somewhat confused. As far as I understand the camera is actually doing the focusing and the lens assembly just acts as a servo. Now I understand that servos can both under and overshoot, and that the amount of over/undershoot may be specific to each lens. Any error related to the flange distance should be specific to the camera body and not to the lens, in my view.
Autofocussing accuracy is a major problem which Canon appears to be addressing by adding micro-autofocussing adjustments to their latest cameras. The recently announced 50D has such adjustment capability and any adjustments can be saved in relation to a specific lens.
As I understand, the alignment of the lens elements in their groups is fixed during manufactuire. If some minor misalgnment has taken place, resulting in the lens being less sharp than another copy from a different manufacturing batch, then nothing can be done later to correct this, so I believe.
This is why, in my view, it is always best to start off with a lens that is sharp. Why waste one's time testing autofocus and flare issues if you subsequently discover the lens is not sharp. If the lens is not sharp, there's little you can do about it, except perhaps always use it at F11. In the case of some ultra-wide angle lenses, they are sometimes not even sharp at F16, in the corners on full frame.
Flare and a lack of precise autofocussing, are definitely concerns. However, it is possible to get around them. If calibration doesn't fix the autofocussing problem, Live View allows for great accuracy of manual focus. Flare can often be stopped by holding a card or hat at the edge of the lenshood, blocking the sun's rays.
My Sigma 15-300 has a repution for being susceptible to flare because of its bulbous, protruding front element. It's the nature of the design. I think the Nikon 14-24 also suffers from a similar problem for the same reasons, but perhaps less intrusive due to better internal coating. Sometimes just holding the palm of one's hand between the direct rays of the sun and the edge of the lenshood is sufficient.
I also wonder if such attributes (apart from autofocus accuracy) which are not reflected in MTF charts, have the same degree of QC variability as resolution.
If a lens has a reputation for bad flare, how would you test different copies of the same model in order to select the one which produced the least flare? Has anyone done such a test and found that there is significant variation in proneness to flare amongst different copies of the same model?