You're confusing diffraction limited aperture (DLA), which is where the effects of diffraction are first barely detectable using a specific digital sensor, with the diffraction cutoff frequency, which is reached at a much smaller aperture. For most cameras it is much smaller than the lens' minimum aperture. For a digital camera, the DLA is determined by pixel pitch and nothing else.
It's quite conceivable that a sharp lens with a high resolution sensor can still exceed the DxO standard for "green"¹ even at apertures well beyond the DLA. It's also the case that the overall "sharpness" of an image of a three-dimensional scene can continue to increase, due to the increase in depth of field, well past the DLA. In order to see any visible effects of diffraction at a sensor's DLA, one must be looking at a large enough magnification to see individual pixels. As the aperture is narrowed, the magnification needed to see the effects of diffraction decreases.
Since diffraction is caused by the oscillation of photons interacting with the edges of the aperture blades, one thing that must also be considered is that diffraction is a result of the actual physical aperture size (and shape of the edges of the blades), not the size of the entrance pupil. When light is refracted by lens elements that bend the path of photons, the width of the oscillation does not change, just as the frequency with which they are vibrating does not change. When the hole they are passing through is smaller, a higher percentage of the photons that pass through will strike the edge of the aperture and be scattered. A 70-200mm f/2.8 lens almost certainly has a larger physical aperture at f/32 than a 24-70mm f/2.8 lens does at f/22, and this will hold at similar lower f-numbers.
Interpreting DxO Mark graphs in terms of diffraction
When looking at DxO Mark graphs depicting lens performance, one must be careful to insure one is properly interpreting what one is seeing in the graph. If a "field map" shows noticeably better performance in the center of the lens than at the edge of the field, the lower performance at the edge is not due to diffraction as much as it is due to other aberrations. The most likely culprits, in rough order of amount of influence:
- A field of focus that is not flat. Refractive lenses naturally demonstrate field curvature. By the way, that's not alway a bad thing. When we use corrective elements to make the field of focus flatter, we never succeed in making it perfectly flat.² At best we wind up with a field of focus shaped sort of like a lasagna noodle. What this means is that when the center of the frame is at best focus, the edge of the lens may be focused a few millimeters in front of or behind the edges of the flat test chart.
- Astigmatism. Astigmatism is the difference between contrast along sagittal and tangential lines. Sagittal lines are like the spokes of a wheel that radiate out from the center of a lens. Tangential lines are perpendicular to the sagittal lines. When the solid lines and the dotted lines shown on a DxO 'Sharpness Profiles' graph diverge from on another, that is an indication of astigmatism. If we look at MTF curves for a lens, the difference between solid and dashed lines shows us the same thing.
- Chromatic aberration. Different wavelengths of light are bent at slightly different angles by the same piece of refractive glass. While we can optically correct for CA to a fairly large extent, it's similar to the correction for field curvature: we never reach a place where all wavelengths of light originating from the same point will converge on the exact same spot in the image projected by a lens. Lenses with elements that bend light more strongly (wider angle lenses) tend to demonstrate more CA than lenses with elements that bend light less strongly (longer focal length lenses) when given the same degree of CA correction.
So what's the best way to see the effects of diffraction using DxO data?
Use the "sharpness profiles" graphs, rather than the "sharpness field maps."
When comparing lens performance at different apertures, look primarily at the lens' center performance on the left edge of the graph. When center sharpness begins to go down as the aperture is narrowed, that is an indication of the first effects of diffraction. It starts very subtly. With your EF 70-200mm f/2.8L IS II at 70mm, notice that f/5.6, at 79.7% acutance in the center, is ever so slightly lower than 81% at f/4. By f/8 there's a little more drop in center sharpness to 78.3%. At f/11 the effect of diffraction has more significantly lowered acutance to 74.4% and by f/32 the effect is devastating at 61.4%.
At the same time, look at the edge performance and see that as the center sharpness drops, the edge performance drops less and the slope of the profile from center to edge becomes flatter. This is due to the increased depth of field as the lens is stopped down and the effect of the difference in focusing distance between the center of the lens and edge of the lens, due to the shape of the field of focus, is minimized. On the edge the increase due to depth of field continues to overcome the effect of diffraction and acutance gets slightly better all the way to f/8 (73%) before it drops significantly at f/11 (71%) and crashes to about 60.6% at f/32.
¹ "Green" does not mean "perfect". It means more than a specific percentage of "use case acutance" is observed. Anything above about 70% will show "green" on the DxO Mark "sharpness field map" charts and the shade of the color is not more yellow than green until about 65%. I can confidently say that There Is No Perfect Lens. Further, There Is No Perfect Lens Test, Either. Get out and shoot!
² Roger Cicala, founder and lens guru at borrowlenses.com, has written a two-part series about the shape of various lens' fields of focus and what they can tell us about a lens' optical alignment: Fun with Field of Focus Part 1 and Fun with Field of Focus II: Copy-to-Copy Variation and Lens Testing