I think the slight upper bass emphasis in the 650s can actually work to our advantage when trying to do post work with them, because it somewhat lessens the headphone bass translation issue when eq'ing the sometimes problematic the 'mud' range. IMO ending up with slightly less than the ideal amount of energy in that region (and below) is a better tradeoff than a little too much. It's easy to fall prey to dialing in a bit of extra bass that sounds good on headphones, so cans with a very slight emphasis there helps to keep the eq decisions slightly 'lean' around that region.
As for the lowest octaves, I can detect the general presence, detail and timbre easier with the 650s than on my big floorstander speakers, but eq and level decisions based on that are suspect simply because of the whole bass headphone thing in general. Still, if I can get the trickier upper bass right, the lower bass often falls into place relatively easily.
The part were I always want to speaker check headphone mixes most closely is the upper mids and treble, because small changes made there always seem more overtly evident to me there than the bass region.
As an aside- I think that one of the bigger challenges in mastering our stuff with headphones isn't actually their frequency response curves (which can be more or less compensated for by 'learning' the sound of your phones and how the decisions made with them translate.. or by personalized headphone eqs, another topic altogether) but dynamics. Headphone listening acts somewhat like a compressor in highlighting micro-dynamic details that aren't nearly as apparent over speakers. That in turn has a big effect on how we hear and make decisions about not only dynamics but also timbre. By that I mean that even if you had a tuned, personalized eq setting to get the perceived response of your prefered phones as close to that of a pair of reference monitors as possible (bass-shaker attached to your seat?), the difference in how we hear dynamics on headphones would still affect your eq choices differently than if you were using speakers.. as well as the other, perhaps more obvious things such as soundstage and the dynamics in dynamic terms alone.