I know DXO scores are pretty arbitrary, but how is the zeiss 6 points higher than the 300/2.8 even though its worse in 4/5 of their "categories"
Because of the difference in speed of the two lenses. Dxo seem to load the 'transmission' score more heavily.
But is a 1.4 lens better than a 2.8 lens per se ?
Well, you see a Score, and you see a bunch of Measurements under it. Logically, you assume the Score represents some sort of summary or synthesis of those Measurememts. But, no...that's DxOMark's Biased Scores for you - BS.
The Lens Score is actually based on 'performance in 150 lux illumination' - the light level of a dimly lit warehouse. So, the Lens Score is determined mainly by the T-stop of the lens...and by the camera on which the lens is tested. Like I said...BS.
I find DXO Mark's measurements OK, as long as you're willing to look past the headlines and read the actual results (and not just the summary table). Like most journalism, it catches audience attention with simplistic headlines, not long lists of qualifications and caveats. There are complicating factors when you're trying to compare the resolution of a normal to a 300mm lens; Photozone comments: "Some may be surprised that all these long tele lenses don't deliver a better performance than many wide-angle lenses. Please note that there's QUITE a bit more space between the test chart with super tele lenses so the longer the focal length the higher is the amount of air diffusion" [http://www.photozone.de/canon-eos/173-canon-ef-400mm-f56-usm-l-lab-test-report--review?start=1]
although this is in relation to the 400mm f/5.6L on APS-C, the principle remains. If I were to criticize DXO Mark's comparison, it would be for making this sort of facile comparison at all. From a photographic point of view comparing a 55mm to a 300mm lens makes about as much sense as comparing a sports car to a tractor.