There's definitely something fishy going on when they test these lenses on Pentax cameras. The following screen grab where the 35/1.4 "A" was tested on K-5 and D7000 show completely different MTF curves, both S&M (pardon),
I don't buy that explanation about getting the most out of a sensor (These two cameras have the same sensor, right?).
These sensors don't have micro-lenses that act differently toward the edge of the sensor, and I am assuming the lens is of a telecentric design, so it's doubtful the sensors would effect the outcome as much as is shown (unless they had the previously said microlenses). Sigma also offers a lens mount conversion service, which doesn't change any optical elements to my knowledge, so I doubt any major optical tweaking takes place when designing for "normal" sensors. (that service may be a con as they may just swap out your lens instead of converting it)