Thanks for your response.
The rod is straight, as the bubble stays inside the circle when the rod is rotated 360 degrees. Rod is SitePro 07-4626-TM with 20-minute bubble vial. I remeasured the rod height, and it is 2.400 m as input into Emlid Flow.
Corrections are from an RS2 with baseline of 19.7 m.
I see your point that the tilt compensator is in fact engaging, as for a tilt of 8 degrees, the error should be more than the indicated 6.0 cm. This is showing that the tilt compensator is engaging but not performing to the level that I would have expected. That is, for a tilt angle of 8 degrees, I would have expected to yield a rover coordinate that is within perhaps 1 cm of the expected coordinate.