Opinions vary on this one.
I set the sub's low pass (and I low cut the mains with an active x-over) just above the highest frequency hump or suckout that I want to EQ away. Basically my approach is to fix the room passively as much as I can, recognizing that below a certain frequency (which varies room to room), passive treatments tend to become ineffective and EQ is the only approach that works (for me, anyway, others may have had different experience here).
Note : EQ is much easier for reducing FR "humps" than it is for filling FR suckouts, so bear that in mind as you decide where to set the x-over point.
I've treated my room with lots of "passive" (including bass-buser type hemholtz devices) treatments, but IME these run out of steam somewhere between 70 ish and 120 ish hz, depending on the room. So I fix what I can passively, set the x-over just where the passive stuff starts to fail and let the EQ do the rest below that frequency.
I set slope, phase etc to get the smoothest response around the x-over point before applying EQ. There are two goals here - smooth overall bass response below the x-over frequency and smoothest response thru the crossover point. The latter helps insure an undetectable "hand-off" from main speaker to subwoofer and is absolutely critical for me. Since there are limited bands of EQ, you may have a judgement call in smoothing overall bass response versus super fine tuning the "hand-off". That is your call on the fly, made only (IME) via trial and error.
Many folks disagree with this approach and prefer a lower x-over point. That has never worked as well for me, but YMMV.
Marty