Looking at one of the last lock sequence in figure 1 attache, we can see that the angular position of the SDB1 bench (SDB1_LC_TX/TY) is pretty stable during the lock sequence, it only drops from time to time probably because no more light reaches the SDB1_B5_QD2 quadrant.
As suggested by Eric this morning, we could imagine updating the setpoints (SDB1_LC_TX/TY_set channels) as some points during the lock sequence, updating with the mean or median value of the channels SDB1_LC_TX/TY over a few seconds (10 sec ?).
We should check that changing the setpoints has no effect on the lock procedure (at first sight, it shouldn't) and think about what would be the best moment in the lock sequence to do that (maybe that is a question for the locking team).