Looking at the DR graph it is quite clear that the camera in reality more or less shoots at base ISO and just digs out the shadows more she ISO is raised. It would be interesting to shoot at 100 ISO and expose like it was 6400, then fix in Lightroom to see if it looks the same as "native" 6400 ISO exposure.
My reading of the DR graph, in conjunction with the ISO sensitivity graph, is that an ETTR exposure at ISO 6400 should have slightly better DR than the same exposure at base ISO. DR at base ISO, at print size, is 14.33EV. At 6 stops shorter exposure, which is equivalent to ISO 6400, DR should be 14.33-6=8.33EV. According to DXO, it's 9.17EV at ISO 6400.
In addition, the ISO at 6400 is slightly less than what it should be, if one starts doubling up from the DXO-measured ISO 74 for base ISO. ISO 6400 should be ISO 4736, but DXO rates it as 4211. Such small differences count for almost nothing, except in this case it tends to confirm that the full difference in DR of 0.84EV (9.17-8.33) might well be apparent. At the very least, taking into account margins of error in the measurements and margins of error in the QC processes of manufacture, I would expect there to be a full 2/3rds of a stop better DR when the ISO 6400 setting is used in preference to underexposing at base ISO.
It would be interesting to see if this is indeed the case in practice. 2/3rds to 3/4ths of a stop difference at such a high ISO is of an order of magnitude similar to the DR difference between the Nikon D3 and the original Canon 5D at high ISO, something which created a lot of exitement at the time.