I see that when I switch the videos to ProFoto RGB in Photoshop, but not in AE, not sure what the difference is there.

interesting thing though - I just thought I'd compare this to a DVCProHD clip from a project I'm working on shot on an HPX500. I get exactly the same results as the 5D. I'm thinking this is simply a result of 8-bit conversion from NTSC-safe YCrCB to RGB - and the d90 is probably not converting to 16-235 so it doesn't display the same errors. None of them show any gaps in LAB.

What container of DVCProHD clip? quicktime(mov)?

Yes, quicktime from P2 using log & transfer in FCP.

I found same problems with any mov file imported in CS3 (h.264, sorenson 3,
except mjpeg)

is 76 a .mov with mjpeg? or an avi like the nikon footage? I know there are some issues with quicktime clipping superwhite values on export from FCP, I'm wondering if this is related.

76 is mov with mjpeg inside
h**p://finalpass.net/videos/MJPEG_test/MJPEG-720p-2MBps.mov

As we see the problem at the moment:
- 5D Mark II writes signal in QuickTime H.264 MOV-file in its full quality.
- Apple QuickTime Player can adequately display these MOV-files.
- But most NLE,s (Edius, Vegas, Premiere, After Affects, ...) perceive these MOV-files in a distorted form (see attached image):
--- Shadows and Highlights is lost (red zone);
--- The subtle gradations is lost (gaps in the histogram).

It is pretty much a known problem 'quicktime gamma shift' (search Google). For example: DV - Columns - Brighter Whites; Richer Colors, Part 1

Either we are waiting for improvements in the NLE. Or need a utility that correctly converts MOV-files in a format suitable for NLE.

The decision has not yet been found.