Iliah Borg, founder of RawDigger, wrote an article about Canon's 5DIV Dual Pixel raw file format and how it actually contains two distinct data sets, one for the main sub frame made up of both dual pixels summed together, and another auxiliary data set made up of just one of dual pixels. What he discovered was that the auxiliary set contained 1 stop of extra highlight headroom not present in the main sub frame. This is very significant because it extends the dynamic range of the camera to over 14 stops! Here are example images showing the differences. There has been lots of talk about this on various other internet blogs.
It would be revolutionary if Adobe could incorporate this find into their converters. Perhaps use the highlights from the auxiliary data set when recovering highlights?