I have often wondered if DxO fudge part of their data, or took shortcuts... it's one of those mysteries of the universe.
Quote: Not the first time that DxOMark's scores completely diverge from our own test results... notably the a7S and 1DX II never looked as clean to us as it did to them."
You have to remember, they shoot in a windowless space with nothing but one incandescent light. Which they claim is some kind of typical photography set up.
My understanding is the reason their values are so crooked is their lack of blue spectrum light in their test environment.
Under blue light, an APS-c sensor can be pushed theoretically to almost 100 MP. Under red light it will struggle to achieve 10 Mp. Here in lies my perceived problem with DxO. By limiting their light sources to red-yellow spectrum, they create an un-natural environment that will create much less resolution than out door shooting conditions. If DxO says your lens is good to 6 MP, I guarantee you it will be on a K-3 good to at least 18 Mp if the light source is blue spectrum. DxO, like any source is only useful if you are understanding exactly what they are talking about. If you're shooting a red flower under normal household lighting, then their evaluation may be dead on. If you're using any full spectrum light source, like oh, the sun, or any white balanced LED or Florescent source, it's all nonsense.
Just my take.