You’ve re-calibrated yet the results are consistently 2/3mm off.
Then you are consistently re-calibrating incorrectly.
![Wink [wink]](https://www.festoolownersgroup.com/Smileys/default/winksmiley.gif)
The splinter guard has no relevance.
Position the parallel guides close to the ends of the board and set the stops to the same increments.
Make the cut (assuming you have taken the slack out of the fit of the saw to the rail)
Measure both ends of the freshly cut piece.
If there is a difference adjust one of the stops.
The last step is likely the problem.
If you have the FlipStop v2, the stop rod has a sliding stop between the fixture and the business end. (TSO calls it the Calibration Memory™ collar)
On the FlipStop that made the cut too wide, loosen the set screw on the movable collar and slide the collar snug against the arm of the FlipStop and retighten the set screw.
Find a shim the exact same thickness as the cut discrepancy. A feeler gauge is handy.
Loosen the set screw in the FlipStop arm and slip the shim between the arm and the collar then retighten that screw.
Pull the shim out and make another cut. The discrepancy should now be only one or two 1/10ths of a mm. In the range where the difference comes from your ability to set the stop flags to the same increment.