There's lots of power left at 3.5 volts.
True, if measuered under adequate load ...
False, if measured idle ...
Li-ion's have a linear discharge curve, it doesn't drop off quickly.
LiIons have a linear discharge curve over almost the whole time, but there is a steep drop-off near the end. At least, if you apply a constant current for discharge. This dropoff shifts with higher loads to lower voltages. That "voltage sag" is the reason, why protection circuits are shutting off at such low voltages. Otherwise you won´t be able to use the full charge with high loads. Negative point of that is, that a low discharge doesn´t trigger protection early enough.
If you're using a high discharge, for li-ion it's anything over 1C, you would be doing the cycle life a good favor by setting the cutoff for about 3.2 volts just to make up for voltage sag and give yourself some wiggle room.
... which reduces the usable capacity highly.
This whole problem concludes to one question which I couldn´t find an clear answer for during all the years here
: How big is the impact of voltage sag onto the critical voltage levels and therefore battery health. You hear all the time that you shouldn´t go below 3.0 volts ... ok, but how bad is it, if I undercut it under load while the battery (idle voltage = state of charge) isn´t nearly discharged, thus would be way above 3.0 volts. According to the LiPo guys at RC every "millisecond" being off the "safe zone" has influence ... at least during loading. But is that also true for LiIon chemistry under load ... and how big is the influence?