Maybe one of these?
http://www.digikey.com/product-detail/en/RLB1314-4R7ML/RLB1314-4R7ML-ND/2561400Claims Q = 100 (at 8MHz, presumably more like 50 at 2MHz).
Awfully large, although at least it isn't very expensive.
Who knows, it might push peak efficiency up to 91, even 92%.
Why is it so important that efficiency be high as possible? You certainly aren't using the best controller if those last six milliwatts are that precious.
I can't imagine many places where thermal would be a problem, at this power level. Perhaps miniature circuitry in a vacuum? You should use a hybrid for that though, much better conduction (bolted to a heatsink for eventual conduction or radiation) than PCB. Or find ways to avoid putting stuff in vacuum in the first place.
If this is battery operated, going from 70% to 90% means going from 7 hours battery life, to 9 hours (assuming this is the only load, also). Two hours gained sound pretty valuable. Going from 90% to 95% only gets you another half hour, for significantly more effort and expense (now you're needing one of those $6/ea super high efficiency controllers that LT makes). Trying to push 99% will require significant design effort, size and cost (you have to do much of it discrete, because no one makes an integrated regulator that efficient), and net only another 24 minutes battery life. Effort that's likely far better worth spending on optimizing the underlying device, which might have even easier gains (e.g., implementing power management settings in all devices, not just the CPU?). Or get a 10% larger battery, which might not even break the mechanical design.
Also, don't forget the leakage spec of the power controller. If you aren't using it at 20mA for 100% of the time, then its average efficiency will be significantly worse. Battery operated devices often achieve big gains for using a relatively crappy (70-80% efficiency?), but extremely low leakage (~uA current consumption between bursts of activity) regulator. Or, consider using ICs that can operate directly from the main supply without reduction.
Engineering is the art of optimization, and there are a great many ways of doing that. Don't lose sight of the overall goal, and all the parts that go into reaching it. Well and truly maximizing efficiency, to the literal, absolute exclusion of any other consideration, is never a positive outcome when it comes to design. Likely, your application is much more diverse and nuanced than this singular constraint, so be careful what you say and think about -- it can mislead us in answering questions, just as well as it can mislead yourself down fruitless paths.
Tim