Having had some down-time to think about this...
I don't think the dynamic range of the camera being increased; that to me seems like a hard-wired limitation of the sensor itself. To my understanding the sensor simply has X amount of DR, say 12-stops for instance, and that never changes. How could it? What DOES change when we use a neutral density filter is how much DR we are allowing to reach the sensor.
Lets say without an ND filter we have a scene that requires 12-stops of DR. This is two-stops more than our theoretical camera body can record. If we then put a three-stop ND filter between the scene and the sensor the overall DR of the scene reaching the sensor is reduced from 12-stops to 9-stops and the latter range, 9-stops, is below the sensors threshold and so can be captured fully in a single frame. Whether the ND filter is graduated or not, I don't think matters; the same principles would seem to me to apply.
.....