The graph posted by Eppie looks better from an FR standpoint. However your waterfall plot shows significant discontinuity right at crossover. That stands out even more in your in your polar map. Now unless you can use very low order crossovers there will be discontinuity at crossover, mainly due to time aberrations. I personally avoid designs with crossovers in that really sensitive 1k to 2.5 K range if possible. I just find they are prone to getting noticed. I suspect the rise in FR is due to what is basically a horn loaded tweeter, that you can call a wave guide if you like. But to be honest that is a horn loaded tweeter and getting rid of that hump is very difficult. It is reminiscent of the FR of the
Dynaudio D21 tweeter.
I used that tweeter in the speaker I won the Stereophile Audio Sound off competition with back in the eighties. I had a really hard time dealing with the FR of that tweeter, but I tamed it. Crossover is 2.5 KHz.
That speaker seems to have a very similar design concept to those Arendals.
If any dealers this way on stock them, I will give them a listen.
The only solution to these crossover problems is DSP in active crossovers.
I do design my speakers to be listened to on axis by the way. I find this gives the best overall room responses, mainly because the off axis sound radiation is more symmetrical in the room I suspect.