The DxO "scores" are in no way to be mistaken for "reality". They are what they are, an end sum of an arbitrarily weighed set of measurement points that in their constitution may (or may not!) in some way be representative of how you as an individual use your camera.
As stated so many times before, you cannot put a "score" on a camera as a complete system for any sort of user average, you have to specify a quite tight definition of the score scope - what it does include and what it doesn't. And it does get tricky when you try to include more than one parameter in a "score".
You may (in my opinion...) for instance put a reasonably real-world relevant score on "low light noise performance". This score will then preferably accurately represent how much noise a camera will show in a shot given an average exposure of exactly "X" lumen seconds per mm2. The measurement "noise per lm*s/mm2" is just a strict definition of what a photographer would call "I want 1/200s at F2.8 in this light".
-But it (the score) will not say anything at all about how well the camera can AF in that same situation. Neither will it say anything about the viewfinder, the fps rate or the color accuracy. The score will be "low light NOISE performance", not "low light performance".
...
But the individual measurement scales and results in themselves that DxO use to build their total "score" are almost beyond reproach. You could possibly fault their presentation mode in some of the results, but not the results in themselves.
The few individuals and organizations that can replicate the accuracy and scientific stringency of their tests all arrive at the same conclusion - there's not much to be said about it except for the fact that DxO are almost 100% repeatable and accurate. The numbers just fit, and anyone measuring the same thing on the same cameras will arrive at the same result. And anyone comparing two cameras out in the real world can - if they adhere to reality in stead of maker-defined bogus ISO values and so on - also visualize exactly what that number means in an image.
So it's up to the reader of the results to make use of the information in a way that he or she can relate to, and here's where most people fall flat on their faces. Unfortunately DxO trip themselves up by trying to apply an overall "score" on a camera, but as far as anyone in the publishing business can tell - you NEED a promise of an end score or other simple graphical or single-digit grading system to get 95% of the readers to even bother starting to read.
If the average readers realizes that they actually might have to think a bit for themselves to get something out of something, they quite often just bugger off somewhere else before even starting to assimilate any real information.
