We are in the digital age that means hardware and software work together one cannot function without the other. You cannot test hardware without optimized software, Canon designed the hardware and they know how to get the most out of it with the software they also created. We are no longer in the film world were you test the hardware alone the software also has to be tested in the digital age. To properly test any camera it should be tested with the software that was written for it no matter how awkward the interface is, that is the only way to determine how good it is. It is irrelevant how many people use ACR since people are wiling to pay for other programs and choose not to use it. It is also designed to work with different manufactures RAW files it may be the jack of all converters but it is a master of none. dpreview needs to get their act together and start testing cameras with their intended software to measure output and compare them to other cameras, they can use ACR later for a comparison of raw converters. Adobe probably would not like it since it may show that ACR needs improvement.
I understand your argument but I see it as flawed. Whatever procedure Dpreview adopts (probably without exception) there will always be disaffected people whose ego is slighted, who don't like the results and who will therefore criticise the methodology and suggest another converter should have been used. This is the very reason why Dpreview have attempted to standardise their procedure by always using ACR if it supports the camera.
If we had a situation where it was widely accepted that the software that shipped with the camera always gave the best results when converting the RAW images from that camera, then your argument would have merit and ACR would probably not exist. Who would want to use ACR if better results could always be obtained from free software that shipped with the camera.
Now, it might well be the case that Canon have lifted its game in recent years with regard to DPP and that the latest versions of DPP have certain advantages over ACR. But one can't assume that such a situation automatically applies to all manufacturers' own converters.
What happens if one particular converter applies a default level of sharpening which one can do nothing about? What happens if another manufacturer's own converter really is poor software and doesn't do as good a job as ACR? Should the hardware then be judged by the performance of the poor software when other software, such as ACR, can do a better job?
In this particular review of the 50D and its comparison with the 40D, it was unfortunate that initially the version of ACR used was a beta. But they corrected that, didn't they? As I understand, the beta version of ACR 4.6 produced a very marginally less detailed result at the extreme pixel-peeping level, didn't it.