I am not making this up, here is why there is more noise in highlights:
http://en.wikipedia.org/wiki/Photon_noise
What you haven't considered is in normal images are processed with gamma scaling that mimimicks human vision and compresses the brightness range of the image towards the brighter end. The result is that highlights are supressed making noise seem a lot less than it really is in those brighter regions. Even with gamma compression part of white band at the top of the image has higher noise than the dark background next to it as measured in software designed to measure that. To do noise measurements you have to use the most out of focus areas (to avoid texture) and areas that are midtone (to avoid large gamma compression and saturation).
Coming back to astronomy, if there were less noise in highlights, this would mean we would be able to detect fainter astronomical objects from the middle of cities than we could from the darkest mountaintops!
Terry