Its relevant because different phones have different resolutions. If benchmarking was done only on native res, people could say "oh well that doesnt matter its because that phone has a lower res and pushes less pixels thats why its faster". It makes it a little more scientific. It would be weird to read [H]ard videocard reviews where every car is ran at a different resolution would it not? Thats why you have offscreen rendering of a standard resolution so you can compare.
True, but in real world use all that matters rendering at the resolution of the phone.