probably used old sensors in the ones they wanted to test bad and everything new on the sponsers.
That makes me think, they should have tried each controller on each sensor. The problem is not each controller has an easy sensor calibration (if any), although that would be another negative point for those controllers. I think this was the ford-related magazine that did that test, I recall it being a big controversy about 1-2 years back. IIRC They used control gases - basicly with X percent of oxygen in each - to test the sensors. They had a big pipe with all the sensors in it and filled the pipe with that gas.
The problem with the results is that any retail part can be bad on occation, even if they have Juan the "scientist" and his brother Julio testing them beforehand. They'd have to pull several random retail samples from different resellers, then drop the best & worst results from each brand. Knowing the standard deviation of each brand is the info I'm most interested in.
BTW AEM users generally have innacurate readings, general malfunctions, and big-time problems when they try to interface their stuff with any PC or EMS. Some people get lucky, but most don't.