Thanks all for the thinking along.

My setup is quite simple. The motherboard terratec codec to a 75W Hafler mosfet amp (with attennuators and dc servo) and then to the drivers. Behringer mic -pointed straight on-axis to the hornmouth- is connected to behringer 8in-2out mixer with some gain, all eq to 0 and then to the codec again. All wiring is simple low L, low C stuff. The terratec codec can do 24/192 out and 16/192 in, not bad! Levels are quite low though:1V rms out and 1V peak-peak in.

So, as you can see no filter and no eq in the chain. I pondered a little on the results and came to the conclusion that I have been integrating reflections as well. The rta is averaged and the imp has a long sequence. So next measurement will be imp with a 3mSecond window, now I only need a meter of free space. 330Hz as a lower boundary is fine with me. I am still a little sceptical wheter this changes the FR a lot. Probably a notch and peak or two, but we'll see. I'll keep the mic close to the hornmouth.

I wondered though if the on-axis position is the best. The center fin of the horn is in front of it and the two adjacent sectors might have some lobing and peaking at certain frequencies even at a close distance. Otoh off axis measurements can probably not be compared to all other material...

One one account the measurements make sense: they sound awfull allright!
I am struggling a little with the correlation with my first auditory inspection. This seems consistent with the measurements, but should that be the case? I mean hearing sensitivity varies with frequency and loundness, right?

Thanks for the cap suggestion in an moment of unattention this may save the phragms!

Thanks, Coen