Ah, so your antennas are already pretty narrow band (and high gain is required), so you could actually do it with TLs and such. And, you sorta-kinda are with the helical resonators, and exactly are with the combline, so there we have it.
So yeah, it depends what the characteristics of the filters are. Ideally, you would measure the system by measuring the radiation from the two antennas, while driving the feedpoint. The filters are correct when the feedpoint SWR is small within each band, and only the intended antenna is radiating.
You can't generally ensure that radiation comes from the correct one, or that the filters will work with the antennas, or with each other, without evaluating the reactances from all these components, across the spectrum.
That is to say, you can't avoid the fact that, you've got complex numbers spinning around, and when you put two and two together, you get yet another complex number, and in general, their result won't be a well-behaved real number (resistance).
There are some strategies to help out, here -- if you use diplexing filters throughout, and terminate the unused ports, then the filter's output impedance becomes a constant resistance. The output port sees 50 ohms (from the terminator) when out-of-band, and 50 ohms (from the receiver) when in-band. Then, because the antenna sees 50 ohms at all frequencies, its response will be the same as when it was tested in a 50 ohm system (say with a VNA and another antenna to measure gain and pattern).
Likewise with the receiver side, the filters can be designed as multi-stage diplexers, so the high, middle and low stop-bands get terminated, while the high and low pass-bands get passed to their respective antennas.
This is probably an over-the-top, belt-and-suspenders method, but it has the least sensitivity to reactive sources and loads.
Now, with instrumentation, you should be able to build the individual filters normally, and determine their stop-band impedance (as measured at the plane where the filters would be joined -- because again, line length matters, including BNC tees and stuff). Once you know this, you can adjust that port (with L or C, resonators, TL segments, traps/stubs, whatever) to get a high impedance, and then it should be okay to connect the filters together. Then, repeat the measurements to verify that they're behaving as expected (and to tweak them further).
The same process (ensuring complementary impedances) applies to the filter-to-antenna connection as well. If nothing else, at least verify that the filter-antenna pair is working as intended.
Then repeat the measurements with the antennas attached to the joined filter pair, and you'll have the full system verified and working.
Your English is quite good
Tim