The only thing I can think of is that the RFI Filter was also used to create a pulse train, which was pushed back into the windings and induced a current into the rotor.
A crude, yet effective creation of a somewhat steady state AC, via segments of RLC network arrays (or chopped via SCRs or FETs).
Alternately, the filter would contain the AC within the Genny and not let it out to plain-old DC devices, so maybe removing the filter placed an extremely low XL or XC (or both) in front of the AC, which resulted in it's major portion flowing externally from the Genny (or reducing the resonant Hz to a very low or high value - out of sync with the genny and/or unable to produce an effective excitation current across the rotor).
Maybe the regulator required some pulse train from the filter for operation? (like a fail safe device? no filter operation - no genny output, attempting to minimize RFI errors on flight equipment???).
What did the filter look like?
A simple Capacitor will pass AC and block DC, so that could be a key item to drive the regulator!
Scott " 35 " Thompson
Just Say NO To Green Eggs And Ham!