I checked with BIOS engineering about this and they tell me that the Performance settings were exposed in the earlier BIOS versions in error, since the processors in these NUCs do not support those features. So, even though the settings were available, any changes that you made were not actually doing anything.
My evidence is empirical that Current/Burst/Sustained are hooked up. Before adjusting these, my two core performance on a specific parallel computation was highly variable, and often barely better than one core. After adjusting these, my two core performance is usually 1.7x one core (expected, considering the overhead of parallelizing. It has rarely dropped to 1.4x, which could be explained other ways. These computations can be held up by the slowest core, virtual cores aren't much more than an illusion, and even though my Ubuntu Server is stripped down, it can be doing other things.)
It occurred to me that one way anyone could determine whether they're experiencing power throttling would be to either lower the Cores multiplier (which starts at 26 but can be set as low as 8) or lower the Host Clock (which starts at 100 but can be set as low as 80 or raised ). One then expects slower timings. If timings suddenly become much more consistent, I would suspect throttling as a culprit before.
Alas, one can set the Host Clock or Cores multiplier (not easily, fiddle with arrow keys and use main keyboard digits, not the numeric keypad) but they do nothing. (At least, "sudo dmidecode" responds to other changes I make, but not these. I'm running timing tests now at Host Clock 80 to confirm there is no effect. No change to core temps.)
Let me reiterate: There is little reason to lower the Host Clock or Cores multiplier in practical use, but these would make for a fantastic diagnostic tool, if one suspects but cannot prove throttling. Too bad they don't appear to be hooked up.