In addition, it appears that there may be additional internal compensation operating over a shorter time scale, as measuring brightness in an orderly fashion from DAC 0-1024 produces a different luminance curve than when the same values are measured in random order. We have yet to determine exactly what causes this. If you interleave each orderly measurement with a period of black and color channel=max, the curve matches the one made from random order pretty well, suggesting that the issue may involve some kind of feedback control with the back-light (kind of like dynamic contrast) that behaves poorly when the max luminance stays relatively constant for a long period of time. More tests are necessary.
One strange 'issue' to know about is that the gamma curve measured over a Display Port is quite different than the one over VGA (ie analogue). The VGA one is squashed downward, so that the upper 10-15% of voltage levels all produce the same (max) luminance. The overall number of luminance steps available is probably the same, so long as you have a high bit-depth device like BITS Plus Plus to drive the VGA signal.
The panel claims to have 10-bit resolution, but as you would expect, it's not quite that; as with all LCDs, there is a significant loss of resolution on the low luminance end, such that you may need to increment 5-10 DAC values before you get an actual luminance change. Roughly speaking, the display is more like 9.7 bits.
Reported by AlanRobinson.