I personally respect the DxO results in general. As a RAW shooter, what they find with most cameras tends to agree well with my practical, real world findings in use of those cameras. However, their GH1 results make me wonder whether their findings may be skewed by manufacturer tampering with RAW data.
The G1 is known to have the same sensor as the E-P1, E-P2, and GF1, and tests identically to the E-P1 and GF1 cameras in DxO analysis. I assume the same will hold true for the E-P2. As an owner of the GH1 and G1, my personal experience is that the GH1 image quality is, on the whole, better than that of the G1. However, I've always suspected that a good part of the difference is caused by Panasonic applying noise filtering at the level of the sensor, such that it is "baked into" the RAW. If my hunch is correct, then DxO is analyzing "cooked RAWs" from the GH1 and "raw RAWs" from the other cameras.
Regardless of whether I am right about the above, DxO has several limitations. For example, the banding which is known to affect GH1 images (especially when underexposed and pushed) does not factor into their analysis at all. I will note, however, that Nik's Dfine is an effective remedy for the GH1 banding.