I imagine because the off-axis curves are individually taken manually. Also, realistically, most dispersion problems like a directivity mismatch in the forward radiating sound would be visible within the 120° slice measured.
When I wrote "the usual +/- 90 degrees," I was referring to Stereophile's usual measurements for bookshelf speakers, ie, they usually measure out to 90 degrees but didn't in this case. It makes comparing the graphs with other speakers they have measured a little difficult as the scales aren't the same.