I like numbers since they help you out, but as you say, the important thing is the end result, after all, FF gives other advantages in image quality instead of just resolution and noise, and when you are not cropping for the same FOV, there is no competition.
When cropping , the noise 'increases'. One needs to compare the same print size equivalent, not 100% crop since this is what gives real life results on print or monitor. From neuroanatomist tests, In real life, the 7D will continue to perform the same as before, but the 5DII will be noiser than usual in this particular situation, but still better than the 7D.
For detail, we need to see the same print size equivalent too, but with the downsampled example, you might not see all the advantages of the more detailed image when viewed on a monitor since both images might be too detailed for the monitor. I think that the upsampled example is a better measurement for detail when viewed on monitor, but then there is the upsampling algorithm step involved in this comparison. The best is to compare it on a print.
on the same theme, using a TC is better than cropping on an FF camera. Unfortunately, there is no data on DXO with TC. It would be interesting to see.