Why are lumagen test patterns measuring differently than AV HD 709 patterns from PS3?

I am using a Lumagen HDQ. When I use it as a signal generator to measure grayscale, I measure a ruler flat 2.2 post calibration. However, when I use the 0-100 IRE window test patterns on AV HD 709, played back via the PS3 (through the same input on the Lumagen that I am using to generate its internal test patterns), I get a much more eratic grayscale result.

In both cases, the same processing appears to be getting applied by the Lumagen. I'm not sure what to make of it.