I think the OP has got a point, but it's not a "peak" we've reached, so much as a "plateau."
It wasn't that long ago that the state-of-the-art Luxeon 1's and 3's LEDs couldn't handle the current you could get outta a pair of Eneloops, much less a single lithium ion. Then we got Crees that could handle an amp and a half, and were small enough to focus well in a good reflector.
But the amount of light per square mm we can get out of that technology has definitely leveled off. To get brighter, you need a bigger LED; SST-50/90, XM-L, lotsa lumens, but the die size makes it difficult to get much throw out of 'em in a pocket-sized light; and since they aren't much more electrically efficient than the older LEDs, more output also means more heat to dissipate. Until somebody advances the state-of-the-art in luminous density and efficiency, that's not gonna change.
On the other hand, those big dies can handle — demand, even — more power than current lithium ion battery technology can deliver in a pocket-sized form factor. Sure, there are IMR cells that can deliver the current, but then runtimes become impractically short. Capacitor technology is just taking off, but capacity is even worse, and cost isn't competitive. So until someone advances the state-of-the-art in battery technology, that's not gonna change, either.
Then there's the UI problem: the vast majority of lights use a single button to power on/off (and most of those are#$%^& reverse clicky, too), as well as change modes. Some use a secondary mode control: a second button, a control ring, a head-twist. None of these options seem to satisfy all needs: head-twisters require two hands; single-button UIs that use multiple clicks to change levels can't be used for signalling, and aren't suitable for military/police tactical use; the combination of a tailcap switch, and a control ring or secondary button at the head, requires changing grips. Years of experimenting with alternate UIs, but still no ideal solution, so we're all fumbling around with inconsistent UIs and assorted buttons and rings and such. But unlike the luminous density, efficiency and battery technology problems, I think this one can be solved, since it doesn't require the same sort of technological leap.