One thing direct drive, or even resistor drive lights, have over regulated lights is their reliability in extreme situations. Unless your light's circuitry is specially potted and sealed, a light that becomes flooded with water will short out. A direct drive light or a resistor drive light will continue to function even if completely flooded with water.
I should point out that heavy potting is standard operating procedure around CPF for lights that aren't expected to be user-servicable (for values of service involving a soldering iron). Malkoffs do it, HDS does it, Nailbender will do it if you ask him to...
Elektrolumens does
not. On the other hand, I can take my Blaster-NG down to individual components in ten minutes, and that's including the time it takes to warm up the soldering iron. If I ever get around to it, one of his 1990s vintage lights is getting an XM-L star.
There's no good reason for
not potting a driver if you're not going to make a light that easy to mod. And some form factors, like the XR cans used early on by McGizmo, managed to be both potted
and great mod hosts.
A second line of thought is that some driver circuits, like the Ra Twisty and the Joule Thief, can actually get
more usable energy from a primary cell than a direct-drive light could, even one without the current-limiting resistor once your battery gets
really dead.
PS: To answer your question, I hate having batteries that are producing usable light, but
not enough. A good boost driver will keep them burning brightly until all the value has been sucked out. Throwing away half the energy in my primaries rankles, and I can't afford to switch everything over to guilt-free, as it would be a substantial (three digit?) up front investment. Also, primaries are
way better for hurricane season down here.