Observed discrepancies in xsuite/xsuite-fluka/sixtrack-fluka results, mainly in TCTs, seem to be solved when removing the initial offset, which as far as I observed, it was applied twice:
In fluka_builder.py, 2. In the xcoll tracking (engine.py)
Now, it is possible to produce the relcol.dat file getting all the collimators (setting True), or producing a list of the desired collimators.
To consider if providing an option of predefined collimators groups would be a good idea. For instance:
start(..,touches=TCP,..) to get all the TCPs