Which site has the more accurate testing protocol

This is for Ted to get his photographic fix.
Photos to Photons ves DXOmark
should DR be based on noise or actual captured images.

eg: DXO the sony a7s sports iso rating 3700 vers sony a7s3 iso 2500 knowing that the sony a7s3 is binned 48 meg to 12 vers a7s 12 large single pixels. Bills site shows a different result.

Hello, Donald: I am biased for Bill Claff and against anything DxO but, in any case, there's an ISO Standard for Dynamic Range, so noise.