I guess I saw that coming, I was typing the following when you posted;

More on that...

A single woof radiates a fairly uniform, round pattern, which narrows with increasing frequency. However, when you put two in close proximity, the pattern becomes an ellipse, or for interpretive purposes, rectangular. The pattern is just the opposite of what your intuition would suggest though, with woofs stacked vertically, the pattern is wider horizontally, and vice versa.

This is borne out in the 9848 directivity plot I linked to. At 500Hz, the pattern is 135 degrees horizontal, and 60 degrees vertical. By the 700Hz XO point, the horizontal has narrowed to 90 degrees, and begins to blend with the 511 horizontal pattern.

So, if you flipped the woofs to side-by-side, the pattern flips too. You can see by the directivity plot how that won't meet the 511 polars at a reasonable frequency, with the woofs doing

xxx
xxx
xxx

and the horn doing

xxxxxxx
xxxxxxx

Yeah, Altec did side-by-sides too, but with a specific application in mind for which it worked. They also incorporated some directivity "steering" (via phase offset) into the design as well.