Also keep in mind that DxO that there is a statistical error. I remember having read a statement from DxO that it is on the order of 5 points. The real-world differences could therefore be much smaller than the numbers suggest.
Can you point to documentation that supports this?
I agree with Jeremy that the overall number isn't as important as the individual areas. One or another may be more important to one person or another. And, in the end, DxO is only one of several 'reputable' review sources. Imaging Resource being one that seems to often get overlooked.