DxO base their scores on RAW files out of the camera, not JPGs.
From what I've read (which could be true, could be nutter conspiracy-theory), is that Nikon de-noise the raw files (even if only slightly) before writing them to the card. So it's not really testing the 'sensor', it's testing the 'sensor and whatever happens to it before it gets written to the card as a raw file'. Call it cheating if you want (i'm sure some do), but at the end of the day we as photographers can only play with what the camera gives us.
As for site-bias, you'll find that everywhere on the net. Go and read Tomshardware.com reviews (and the user-comments) for the last 10 years. There were always allegations of bias towards AMD in the Athlon vs P4 era, then more allegations of Intel-bias in the years after of Athlon vs Core2. Never mind that the chips they were allegedly 'biased towards' were actually the best chips out there, you just can't argue with fanboys on the internet.
At the end of the day, make your own mind up, look at the graphs not the text. I'm not going to ditch my camera and lenses for an extra 0.5 bit depth or whatever, I don't think many people would. Those who would must have some really demanding customers, or have way too high standards imho...