Today I’ve finally got the all missing parts to complete my two prototype boards.
I used my hackish rtlsdr + noise source + qspectrumanalyzer setup to get a rough idea about its characteristic.
Since the combination of noise source and rtlsdr does not really produce an even spectrum over the 420MHz span, I took some baseline measurements and used some slightly questionable math to straighten it out.
So I still need a proper measurement with proper equipment.
It does look a lot like the simulation, though.
Most importantly the path between the two inputs has a high attenuation all the way through the spectrum.
Looks like it could work.