This test was run on a Nexus One, which was released in 2010. I'd be curious to see if they got the same results on a newer phone. I'd be willing to bet that this is mostly because ads require more network requests, which aren't as efficient on newer phones.
Something smells of a researcher cherry-picking a result that shows off their tool's results in the best possible light. I suppose I don't have any way to prove that, but I have to ask why the researcher didn't try the test on another phone.
A Nexus One is a very convenient target for research because it is (a) sold without ties to a carrier or any carrier-specific junk, and (b) flashable with any specific system image you want, including custom AOSP builds without doing any other work (the hardware profile for the Nexus One is shipped with AOSP).
Something smells of a researcher cherry-picking a result that shows off their tool's results in the best possible light. I suppose I don't have any way to prove that, but I have to ask why the researcher didn't try the test on another phone.