Marc,
your 5D2 beats your 5D by at least 1/2 EV in the darkest regions @ ISO 100 (I used one of your earlier posted 5D raw files). This is on a per pixel basis, add the extra pixels!
Do you still have the color checker card? If you make a serie of raw shots with each ISO, up to 3200 (save the higher ones, they are fake), very evenly lit, 3 EV underexposed, then I can measure and compare their noise/DR behaviour. IIRC, you had converted the 5D in IR, right? That does not matter, the noise characteristics is not affected by that.