That's good shout. I suspect they may have had multiple sources for the balance, and thus the shim was fitted in some cases, to allow them to compensate for the 0.1mm or so difference between shaft lengths.
The shim can be source of problems though as it can act as a pivot point if it is not fitted in entirely the right spot, and this could introduce errors in the balance alignment, and it does introduce the possibility of either forgetting to re-fit it, and screwing the balance down hard on to the jewels, thus damaging the pivot(s) and/or the jewel(s), or fitting it when not needed if a different "compatible" replacement balance is fitted which has a shorter shaft, and then getting in to head scratching mode when the thing rattles around like a pea in a biscuit barrel.
Thanks for the answers. A more detailed description first:
I reassembled, oiled the pallets with 941 and let the watch run for 48 hours dial up.
Then I did the poising and after a full wind the first measurement on the timegrapher with result +2/0/+2.
Then I wore the watch (normal day in the company) for 24 h. Then after a full wind second measurement with result +9/+7/+9.
So maybe the mistake was to let the watch run static in dial up position instead of "shaking" it while wearing? Or the period of 48 hours was too short? Right now the watch is on my wrist and I am observing the "real" gain compared to the atomic clock.