Rich said: "I just don't think their test samples are dirty enough that they could tell the difference."
I think, from CR's description of their own tests, that there are actually a couple of points where their tests can fail, and probably do.
First, they use swatches that are uniformly dirty with special stains -- supposedly, they attach several of each stain (8? 10? I forget or it may have changed over the years) to white cloths and wash it. The problem, from my perspective, is that even if they use several different stains, it's still true that a few dozen swatches do not really make a full load, it probably still has less total dirt than a load would have. It's very different than testing the different detergents on our home, where the loads/stains tend to be similar from week to week and you observe the performance in raw form. So, as Rich is saying, it's possible that the dose of detergent is actually overpowering the amount of dirt they have, but the same dose for a full load of real laundry will leave some dirt behind.
Second, they use a fancy machine that shines light against the samples and measures how much light is reflected, and what colors, to tell how clean the swatch is. Over the years I have used plenty of different detergents, cheap and expensive, and I have often noticed that some of them, often the cheap ones, but not exclusively, tend to give the impression that the laundry is clean -- sometimes it looks like even chocolate milk stains have been cleaned -- but then when you look at light through the fabric you see the stain is still there. I have certainly had the experience of using a couple of detergents that CR said one was fabulous and the other cleaned OK but didn't have as nice whites, and find out that the one that was supposed to be fabulous actually left the chocolate milk stain in there hidden by the optical brighteners, while the other one did not actually look as white but was completely clean. Manufactures quickly learn what the testing procedures are and live down to it quite happily, unfortunately.
You can bet that as soon as they change the testing procedures to shine light through the fabric, some detergent makers will up the optical brighteners to try to hide the stain even more and some others will probably up the amount of enzymes to more fully remove the stains, or possibly a combination of both. And if CR starts testing full loads instead of just swatches, we'll probably witness the disappearance of "HE" vs. "non-HE" detergents, because it will be in the manufacturers' best interest to make the thing as non-sudsing as possible so we can put as much detergent as we need in the washers to get it to clean the laundry and rinse as cleanly as possible.