A well written article, Bob. Just wondering about non-linear sensitivity in the hightlights. You mentioned this in the dynamic range section.
From what I have read on the Internet the D800 seems to hold more details in the highlights than the 5D3. I do not know if this is true or not but sites like dpreview has made claims about this saying you can do a better recovery with the Nikon.
I did not know that digital cameras could have a non-linear sensitivity in the highlights. Black and white film hold a lot of highlight details, I remember, but digital has never been that good when in comes to highlights.
CMOS sensors are inherently linear and there really isn't anything you can do about that if you have only one type of pixel. Fuji have used two different pixel types in their SuperCCD SR sensors to extend dynamic range, but neither Canon nor Nikon do that.
I suspect that "extended" highlight dynamic range is done via digital tricks. For example Canon have Highlight Tone Priority. Canon have never talked about exactly what they do in this mode, but since it sets a minimum ISO of 200, I suspect that in fact it underexposes the image, then does digital compensation to bring up the mid tones and shadows. This would result in an apparent expansion of the highlight "headroom", effectively creating a non linear response curve via some sort of software/hardware function. You could do this either in software, or by making the amplifiers that condition the signal from the pixel (before the A/D step) with a non-linear response curve.