I must confess I was quite under-whelmed by the article as well. I noticed many of the same things mentioned above. In addition, I'm afraid I consider the article methodologically unsound.

A comparison of two developers is a complicated thing to do, because as Jay pointed out, there are so many dimensions to compare them on. To make menaingful comparisons, you typically want to control everything else except the dimensions of interest in a particular test. For example, if you want to test the film speed yielded by two developers, you have to develop them to the same CI etc. What this means is that a comprehensive comparison of two developers requires multiple tests across a variety of fronts. A really rigorous comparison would involve multiple measurements to make sure small differences are significant and not just due to random error, but with the reseach labs at Kodak gutted, I doubt anyone is going to have the patience/money to do that. At the least, one would require some level of experimental control. That would give us a theoretical understanding of the differences between the two developers.

Unfortunately, not everyone is comfortable with interpreting theoretical results and translating them into practical situations. That means some folks prefer to make pictures with the two developers and evaluate the results. Whether those results can show up in magazine reproductions is most unlikley. The biggest problem with this approach is the fact that there are interactions with subject matter. That is, one developer might naturally do much better with certain subjects than another and if that is the subject you shoot, you have skewed the results of your tests right off the bat, which is why I personally think this is methodologically not sound. It is absolutely fine if it is one individual trying to choose a developer to work with because s/he is interested in choosing a developer for his/her subjects. That is, the conclusions s/he draws are indeed constrained to the subjects shot. But in this case, there are generalizations being made that cannot and should not be made. Having used both developers a fair bit, I am highly dubious whether there are substantial differences between PMK and Pyrocat on the reported dimensions.

Cheers, DJ