The RFI Filters are working correctly, and acting normally, to filter out what is figured by the filter to be interference.
Those parameters are the design points.
These AC Line Filters are most likely built as a typical PI filter with a common mode Inductor on both lines - instead of a single Inductor on one line.
At the line side of the Inductor(s), two Capacitors will be connected in series with each other, but in parallel to the Inductor.
These are connected to the AC two wire circuit at the input.
At the center of the series connected Capacitors, there is a tapped lead - which goes to the EGC (tap driven to ground). That's where the leakage current is coming from (or going to).
On the Load side of the Inductor(s), there is a single Capacitor in parallel. It connects directly to the AC two wire circuit, but has no "Ground Tap" like the input section does.
Some more elaborate Filters may have several "PI" stages - therefore may have more than one "Ground Tap Point".
These Filters will be low pass filters. They probably are designed for passing Frequencies upto 400HZ, and "Dumping" (rejecting) Frequencies >400HZ.
The higher Frequencies will be dumped to the Equipment Ground/Equipment Grounding Conductor, via the center tap on the input filter section.
This may sound funny, but an effective fix for this scenario could be done by using an Isolated Equipment Grounding Conductor that connects only to the filters' EGC terminal, then terminates in a location which can be detected by the leakage monitor, or possibly passes the monitor.
Other than this, the filters might need to be exchanged for more simpler types which do not dump to the grounding conductor.
I can think of 3 things being filtered out to ground, by these filters -
[*]Carrier Current signals,
[*]Actual Radio Frequency Interferences.
Curious to hear what the Designing EE has to say.