There is a popular "explanation" for their results...Canon provided them with a hand-picked copy
Canon would have had to custom-build one with a different optical formula...I seriously doubt that Canon could find a copy where the 17-40 at f/4 delivers corner sharpness comprarable to the center, and is better overall at f/4 than the 16-35 II at f/8.
Note that when looking at the 16-35 II, it's actually sharper at f/2.8 than at f/8 - also pretty unlikely. DxO's measurements of 17-40 show that it doesn't get sharper when stopping down to f/8, because apparently it's already very sharp wide open. Yeah, right...
DXO is such a reliable site...
Like I said, their Scores make sense if you always shoot in dimly lit warehouses. In 150 lux illumination, the better high ISO performance and dynamic range (you're going to boost those shadows, right?) of the D3x make the Zeiss 21/2.8 a better lens, and that clearly accounts for it's higher score.
Oh, wait…you thought that the Scores were logically derived from the Measurements? Silly, silly tron.
... still, after getting the ff I can say that the lens is indeed excellent unless you're looking for corner sharpness wide open, f2.8 speed (imho seldom necessary at this focal length, and these are not very bokeh'ish anyway)
The 16-35L II doesn't deliver corner sharpness wide open, either. If you want that, you need to get the TS-E 17L or TS-E 24L II (and don't tilt/shift them too much).
Personally, I have found that the extra stop of the f/2.8 lens helps for low light shooting (and with 'normal' subject distances, you don't run into issues with too shallow a DoF). But I really think that extra stop of light is the only reason to pick the 16-35 II over the 17-40 - for real-world use especially stopped down a bit, there's not a significant difference between the two (or if you believe DxO, the 17-40 is actually much better