I ended up ordering some MC34063A chips with the appropriate inductors for the series approach, and some 60V 8A 0.18 ohm r on mosfets to try the parallel approach. Since all the parts for both with extras were like $5 I figured I'd leave my options open. We'll see how it works out.
I came in late but here's another tool you might work into your options. It's basically a linear regulator in current regulating configuration with an enable pin X3 channels. The potential advantage is that it prevents LED dimming with battery voltage sag as long as you design in enough headroom. Of course, voltage sag won't be a problem if you use the boost converter.
Speaking of efficiency no one mentioned that MC34063A is not 100% efficient, depends of the voltages it can go as low as 60% efficiency.
Also as far as I know, the value of inductor calculated is the MINIMUM value for the given ripple.
What about building some sort of constant current driver ?
That looks pretty good for this, I would have put one into my mouser order, but now I need to wait till I'm ordering enough to justify the shipping. It sounds like it can keep the current constant over a range up to 15V, so that would be about 5 LEDs per string, which would actually work really well if I want to make a set with a lot of separately controlled lights. I also like that it's a pretty simple circuit. Thanks for telling me about it.
Mouser let me add the parts to the order! pretty nice of them. So now I can try 3 approaches. This also gives me a chance to try surface mount components, which I've been wanting to do since I've gotten the capability to etch my own boards a few weeks ago.