users of these sites, including DXO should be aware that they do NOT test lenses, and do not have equipment capable of testing lenses.
What the sites do test is a lens / camera combination. What this means to those who read the results carefully, and they tell you this, is that the results only apply to a lens used on that exact camera model and not on others.
I've found that test results on the different sites for the same camera / lens combination are usually very close, but if a lens is tested on a Canon 30D on one site, but a Canon 7D on another, the 7D should always have a higher resolution.
This is due to the fact that the system resolution is the product of the mtf of the individual components, and that includes lens, sensor, in-camera processor, software on your computer, the test chart, lighting, and even a monitor used to display the image.
Its no wonder that testers using different cameras and equipment get different results. Lenstip, for example, admits that they get different test results from others because they use different charts, among other things.
However, if you pay attention and read the test procedures, and compare ones that use the same camera / lens combination, the results are almost always close.