Describe the bug
When importing an RTPlan (with MLCS) and RTDose from Eclipse, recalculating the RTPlan within matRad will lead to differences in the dose compared to the imported one. The recalculation appears to be correct for open fields, but not for fields with MLCs. We have performed our experiments for a FFF 6MV kernel in using a cubic water phantom.

To Reproduce
Steps to reproduce the behavior:

1. Generate a machine specific pencil beam kernel using https://github.com/e0404/photonPencilBeamKernelCalc
2. In Eclipse (or presumably another TPS), create an IMRT plan (single or multiple beams). Add MLCs (easiest to just fit the MLCs to the target).
3. Calculate the dose in Eclipse
4. Export the dose and plan from Eclipse and import into matRad.
5. Recalculate the dose in matRad using the custom kernel and compare to the imported plan.

Expected behavior
For a water phantom, we expected the recalculated doses to be very similar to the imported doses.

Screenshots
1-beam IMRT, 10x10cm field, open field
DVH from imported dose

DVH from recalculated dose

1-beam IMRT, 10x10cm field, MLC fit to target
Gamma pass fraction at 2%/2mm over the phantom = 96.79%
DVH from imported dose

DVH from recalculated dose

2-beam IMRT, 10x10cm field, MLC fit to target
Gamma pass fraction at 2%/2mm over the phantom = 94.69%
DVH from imported Dose

DVH from recalculated dose

Desktop (please complete the following information):

• OS: [e.g. Windows]
• Environment & Version [e.g. Matlab 2023a]
• matRad master branch

Additional context
When recalculating the dose, we are using pln.propDoseCalc.useCustomPrimaryPhotonFluence = true;, but this does not seem to have any effect since the dij calculation is very fast.