There have been many thoughts on this, in the past.
Here's the way that I did it:
(I started using a discrete test tone CD too, then migrated to my PC with
www.etfacoustic.com software. Much faster.

) Crossover engaged the whole time.
1) Plot my sub by itself.
2) Plot the mains by themselves.
3) Do all the equalizing with only my sub.
4) With both sub and mains, I plot through the crossover freq and set the phase for the maximum SPL with the eq. (That phase relationship can and will change with an eq in the loop.)
5) Then I look for that total sub+mains response with and without the EQ and compare to what the sub looked like alone with and without the eq.
6) Make adjustments from there.
I always did the sub by itself first, because it's a lot easier to see the effects of the changes. But. Bottom line is the total freq response of the subs and mains.
It is fun, isn't it?
Same here: BFD.