Could that discontinuity at around 700Hz in the horizontal contour plot be the result of baffle step?
It depends a bit on how you define "baffle step". If you consider "baffle step" only as a transition region where radiation goes from omni-directional to directional, then this is not sufficient to describe the on and off-axis FR.
If you consider the baffle step as part of the edge and cabinet diffraction of a loudspeaker, then you can explain the effects on the frequency response and the radiation of the speaker.
What is often forgotten, the cabinet depth also plays a role and represents a parameter with which also influence the radiation pattern of a loudspeaker - depending on the cabinet shape and size, the influence can be very significant (therefore, a 2D baffle simulation is not sufficient to simulate the effects).
I had already simulated the "old" version of the speaker in the other Sointuva thread. If you look at the simulations for an "ideal" woofer (in a closed cabinet, with dimensions like Sointuva), you can see the influences of the cabinet on the radiation very nicely.
Horizontal normalized sonogram (+-180°) and normalized FR (+-90°) Sointuva woofer:
The simulation shows a radiation pattern for the woofer that is comparable to Erin's measurements of the Sointuva up to near the crossover frequency.
One can nicely see that there are several times of widening and narrowing of the radiation, which is due to the interference of the woofer with the various "pseudo-sound sources" of the edges.
Vertical normalized sonogram (+-180°) and normalized FR (+-90°) up and down Sointuva woofer:
In the vertical direction, the woofer is arranged asymmetrically on the baffle, which results in a radiation pattern that is significantly different from the horizontal one.
Edge and cabinet diffraction shows of course every speaker, by appropriate design you can control the radiation pattern and must decide what compromises you take.