It's very doubtful the wider throat diameter may play even a very small part. The optical formula for both versions of the lens are the same, so the Canon lens would theoretically perform the same even if Canon's EF mount had a throat diameter as small as Nikon's F mount. In other words, the Sigma design does not attempt to exploit the possible advantages the 54mm throat diameter of the Canon EF mount offers over the 44mm throat diameter of the Nikon F mount.

The most significant variable between the two tests is that the Canon EOS 5Ds R has a 50 MP sensor and the Nikon D800e has a 36MP resolution. If you play with the tests at DxO, you see very quickly that the resolution of the camera can have a dramatic effect on the performance of the same lens. Substitute the 30MP Canon EOS 5D Mark IV for the 50MP 5Ds R and the performance of presumably the same copy of the same lens is lower in the test results. The 36 MP Nikon D800E is closer to the 30 MP 5D Mark IV in terms of resolution than it is to the 50 MP 5Ds R.

The most likely explanation for the difference in the results is in copy-to-copy variation. It could probably just as easily have been a Canon mount lens with less stellar corner performance and a Nikon mount lens with better corner performance as tested at DxO.

That's one reason Roger Cicala, the founder and lens guru at lensrentals.com, usually likes to test at least ten copies of a lens before publishing any test results. You might be surprised at the variations he sees from copy-to-copy of even some fairly high end lenses. Zoom lenses are even more complex and usually have a higher amount of variability, but prime lenses with more lens elements can also demonstrate increased deviation from one copy to another. With 15 elements in 11 groups, the Sigma 24mm f/1.4 ART is certainly a candidate for fairly high copy-to-copy variation for a prime lens.

In his blog entry regarding the Sigma 24mm f.1.4 ART, Roger says this in his summary at the end:

More importantly with any 24mm f/1.4 lens, try as you will, unless you really stop the lens way down you aren’t going to get a flat field of focus.

In another blog entry about measuring lens variance, Roger specifically uses several 24mm f/1.4 lenses to demonstrate his methodology and how he presents the results. He explains why he chose several 24mm f/1.4 lenses:

We specifically chose 24mm f/1.4 lenses for this example because they are complicated and are very difficult to assemble consistently.