You should pay attention to the testing conditions to get the correct interpretation. We are talking about differences within the component variability here because the exact same circuitry and processing are used in the FM band for both versions. Different units from different batches will still have very close performance but the inequality can change the other way around for this specific test.
The test mostly shows that both the HF+ and Discovery behave much better with very strong signals than older generation tuners.
Missing this point is like comparing A=1.0001 and B=0.9999 to C=0.5 then coming up with the conclusion that A and B are different while your uncertainty is 0.01. Of course, they will be different, but that's not the point.
As Leif said in a simpler way (but didn't demonstrate it,) the performance with actual FM signals with the standard channel separation instead of an unmodulated tone at close range should be the virtually the same.
Leif: Maybe this can be explained better for the non tech-extremists in a different video?