I did some deep diving on the interewebs this weekend, mostly threads on other forums (many I never knew existed) and some of the sources they led to. Turns out there's a good amount of research confirming my suspicions above about many factors confounding corrections. It seems hops alone can have an effect nearly as significant as alcohol, as can different grists, let alone all the other compounds in beer. Much more research would be needed if it's even possible to develop reliable corrections, and the pros aren't going to do it, they don't use refractometers. We're just trying to use an instrument to measure something it wasn't designed to measure. (No I don't have all the references, you've got Google if you care.)
This looked good to me on a limited number of samples, but the more I've racked up, it looks like none of these formulas really works, which now doesn't surprise me. If I have all the formulas in front of me (and thanks to Big Monk, I do!) I can find one of the half dozen that's fairly close. But that just means that there's a one in six chance I can squint really hard and pretend it looks like something's working, not that it is.
So for my part, I'm back to the position that refractometers are useful pre-boil only, and I establish a correction factor for each individual batch using the saccharometer OG. Even Terrill has said (referenced somewhere in this thread) that if you actually want to know FG (or for that matter when to spund) you really need a hydrometer.
So finally, if you're of the school of "close enough is close enough, I don't really need to know," why do you need any instrument at all?
When it stops bubbling, its done!