The graph I showed above is the raw, unfiltered freq response for the ER18 midwoofer. Of course, it will look different once Dennis's crossover filter is added. I don't have that graph. Dennis is highly aware of the imaging created by good off-axis performance, especially in the crossover range, and all his designs take that into account.
Here is a graph from the Salk SongTower, a speaker similar to the ER18 MTM. The differences are it uses 5¼" midwoofers, a ¾" dome tweeter, and a crossover frequency a little under 2.5 kHz. The ER18 MTM response will strongly resemble this.
Two traces are shown, on-axis and 60° off-axis. Below 1 kHz, both traces are close and run parallel to each other. Between 1 and 2 kHz, the 60° trace begins falling off a bit, but the difference is still less than 3 dB. Between 2 and 3 kHz, you can see the two traces get closer and coincide at the crossover frequency. The 60° tweeter response stays close to the on-axis response until it gets above 10 kHz.
For speakers, horizontal dispersion matters more than vertical dispersion. What determines a driver's horizontal dispersion is not the shape but the diameter (for a round driver), or horizontal size (for a ribbon tweeter).
If the sound wavelength is larger than the diameter of the driver, the sound disperses widely. A driver begins to beam as the sound frequency gets high enough for the wavelength to be similar to its diameter. As the wavelength gets shorter, the beaming gets more pronounced. That's the theory, the real measurements can differ.
To convert the size in inches to sound wavelength:
Sound travels at 1125 feet/second, or 13,500 inches/second. Divide speed of sound (inches/second) by frequency (Hz) to get wavelength (inches).
13,500 inches/sec ÷ 13,500 Hz = 1 inch
In theory a 1" dome tweeter disperses sound well as long as its lower than 13.5 kHz. But all speaker makers measure this, and in practice beaming usually begins a bit lower than theory predicts.