In another post, @TB_RTK mentions that weather could be one source of the Phase Center being offset. So, if you make your measurements in a limited time-window, meaning the weather is also the same, then that error would be close to zero.
For the Z-error, in a relative accuracy use case, you can also check this: Evaluating relative repeatable precision of the Reach RS+
Among the point collected here, I have a RMS in Z-axis of 1.6 mm (on controlpoints, checkpoints being 3.6 mm). From that I’d say that the relative accuracy is extremely good.
Now, for the absolute accuracy, I think the phase center is the least of our problems. Things such as baseline, observation time, and pure mechanical issues are much larger sourcees of error.