Well, for one, they don't clearly define the criteria for better vs. worse. The conclusion points out some ways in which the 'Boys were more impressive than the Pats, but they arbitrarily pick other criteria as more important and decide Pats > Cowboys based on that. That's not objective analysis, it's subjective opinion based on cherry-picked quantitative data.
Their "Quality Opponents" measure is very, very problematic. It only looks at opponent records, regardless of quality of their competition, and doesn't include potentially balancing factors, such as losses vs. poor opponents. A 2-12 team could be ranked higher than a 10-4 team in that system.
There are no controls whatsoever in the measures they use to compare the '92 Cowboys vs. the '04 Patriots, even though more than a decade lies between them. As a data analyst/statistician, the first thing that jumps out is that ALL numbers are higher with the '04 Pats...yards gained AND allowed, pass/rush gained and allowed, etc. Also, despite the fact that the Pats YPG was higher than the Boys, they were ranked lower. This suggests strongly that yardage was up league-wide in 2004 compared to 1994, so comparing straight-up numbers is not going to give a helpful measure of relative ability. (If that's even what they're looking for; they don't say)
It's fine if you want to say "based on these criteria, team X was better," but one could just as easily pick other measures, insert some statistical controls, and demonstrate the opposite. In a solid statistical study, you define your criteria ahead of time, and you use statistical controls to clean up the "noise" in the data.
