>> At the beginning the output is around zero because the loops haven’t locked yet. They could run the PLL on the signal backwards. This would provide lock at

>> At the beginning the output is around zero because the loops haven’t locked yet. They could run the PLL on the signal backwards. This would provide lock at the beginning. Combining that phase measurement with the forward one might also reduce phase error in general.