See my full results here, including a table with my dynamic range measurement for each camera at each ISO setting:
I shot this target with all cameras and settings, and present here the ISO 200 results at the aperture that is barely showing usable shadows in each camera. If I close the aperture one stop from this, that deep shadow becomes noisy black with no discernible detail (or at least something that I deemed "not usable").
What you see here is a 1280x720 100% crop from the original image delivered by the camera. I cropped it this way to avoid vimeo from washing out the original resolution. Uploading a 720p file should preserve detail better.
In the H.264 shots, that shadow is a lot more muddy than in all the others, as the codec eats up the detail pretty badly in the shadows.
As for the highlights, in each shot I have marked with a red dot the brightest grey patch that I consider "not blown up". Open the aperture one stop, and it becomes 100% white.
In the Pocket-ProRes, it may look like pure white, but it is actually a bit darker than that, and since this codec is 10-bit, it should be possible to bring it down.
All images have been graded for maximum DR, with heavy slide-pushing in ACR for the RAW images, and with the 16-235 out put levels applied to the NEX footage (you can see at the end how it looks if you don't apply that; it's even worse on real subjects and, in particular, skin).
The 5D2-H.264 is using Prolost Flat (Neutral dialed down; it was a borrowed camera and didn't have Flaat installed), the NEX is using Flaat, and all the RAW images were developed in ACR with Flaat settings.