I am not going to put much faith on that test. First, both lenses need to be taken at the same focal length. Second, they should use a full frame so corner sharpness can be determined. Who legitimately tests an L series lens on a cropped sensor anyways...?
As I pointed out in the other thread where this 'review' was mentioned:
They are only looking at the resolution in center of the lens (my subjects usually aren't right there, are yours?). Also, I don't understand the rationale for summarizing the results as an average of the wide and tele ends or an average of the values from wide open (which varies by lens) to f/16.
Moreover, looking over the tests on that site, I see several results that disagree with many other testing sites (e.g. the 300mm f/4L IS scoring much higher than the 70-200mm f/4L IS, which is quite the opposite of data on TDP, PZ, etc., even when only looking at center sharpness).
So, they're reporting just one measure among many that are important to lens performance, and calling that one measurement a 'review' is an understatement, IMO. Add in the questionable accuracy of their methods based on other lens tests (and in this case, a direct contradiction of Canon's MTF charts), and I'm certainly going to defer judgment on the lens until we see some real reviews (although it's an academic issue for me, since like you, I have no real interest in the lens for my needs). Those reviews may turn out negative, too, time will tell. But the MTF charts for the 24-70/2.8 II were also impressive, and the performance tests were consistent with that. But then again, there were bad copies of the 24-70/2.8 II (TDP tested 4 of them to get a good one), and if the same problems are plaguing the 24-70/4 IS, that may be a factor in poor reviews as well.