You're right to be careful about the results and the comparison. I don't think we as stakeholders of the Eclipse IDE should (nor will) trust them blindly and take them as a reason for not improving performances.
But still, it's comparing what is Eclipse and what is IntelliJ for an external developer; even if they are different packages, this comparison is the natural one made by a user without any attempt to create a bias (I hope), so it's a fair one worth sharing with a wider audience. Somehow, it's normal that the way we package things and deliver them to users and the way user chose a package are also part of the performance comparison.
Now, technically, I'm almost sure it's not a Maven project, because as we all know, m2e is creating much more that 2MB/s ;)