I can only state what I stated before, in the hope that it sinks in. The numbers when comparing are meaningless, the bar heights in relation to the graph height are not all that meaningless. The numbers are higher for the A7r tests because of a higher resolution sensor, not because of a significantly sharper lens.
The aim for PZ is to test lenses, not camera/lens combinations. That is why the higher numbers do not reflect in higher ratings.
Lens rentals often makes odd mistakes in logic. If they what to show how sharp that 35mm f2.8 is, why do they not do the obvious (which is: to put that Nikkor 35mm f1.4 and/or 50mm f2 Zeiss also on the A7r)? THEN you can say something about the lens. The AA-filterless D800E is not really AA-filterless (Nikon says it has two which cancel eachother out). It shows in images. It is not like there is no way to mount those lenses on that A7r, after all.
Anyhow, I hope you get the point. PZ scores lenses, not camera/lens combinations. The bar heights (with disregard of the particular numbers) more or less show how good a lens is, sharpness wise. Look at the "excellent / very good / good / fair / poor" markings instead of pure numbers.
Well I thinik this is fallacy in your statement in that it takes a good lens to produce high resolution on a dense sensor; but having said that a good lens cannot produce a high resolution on a low density sensor.
This is one thing that always bugged me about lens testing. They test with the camera so it is system testing; kind of wish there was an easy way to test without the camera to determine the quality of the lens (though then we run into the crap with interaction between the angle of the light and the micro lenses on the sensor as well as other issues). Hum. Oh well.
I can never place the DXO "colours". So I end up disregarding their "findings" due to the weird nature of them.