Interesting. I didn't know that Roger quote the sensor temperature. My D800's dark current
measurements are referring to ambient temperature.
On my 5D2, the sensor temperature is typically 10 to 14 deg C higher than the ambient.
Let's give 7D2 some advantage and say its sensor temperature is only 8 deg C higher
than the ambient. After this adjustment, according to Roger's graph, at an ambient
temperatures 6 deg C (14 deg C sensor), the dark rate is roughly 0.02 to 0.03 e- per
pixel per second. At an ambient temperature of 25 deg C, the dark rate is roughly
0.4 e- per pixel per second. Translating these to dark rates per surface area, it is
1350 e- per second per square-mm at 6 deg C, and 21600 e- per second per square-mm
at 25 deg C. D800's numbers are 1400 at 6 deg C and 12803 at 25 deg C (ambient).
They are basically on par with each other. (D800 and 7D2 have similar pixel sizes,
4.8 um vs. 4.3 um. So even if we just look at dark rate per pixel rather than per
square-mm, the results won't be too different.)
On the other hand, Roger's measurement method is robust against what Canon did
to alter the raw image. (He is not really measuring dark rates. He measures dark
noise, which is what we really care about.) You can trust his measurements.