Really like the effort to produce actual test data. This will prove to be useful to many.
One caution: The protection circuits of the two stacked cells will be the deciding factor for when this chip shuts down. Efficiency falling off won't be as important as input current going up as it tries to output the 5 volts. As input voltage goes down, the chip will draw more current until the input current limiter stops it. As the input voltage falls, the current from the batteries will go up. If the protection circuit cuts off, or the chip stops, protection circuit comes back on, chip starts back up, etc. etc. If output is set lower, like 4.7, then it would probably work well all the way down to battery cutoff. If output is set higher (like 5.5 volts) then the chip won't be able to supply regulated output when the battery voltage goes down and the chip will shut down early because output is below set point.
That you can get just over 90% is great.
one other note for people wanting to try this type of circuit:::
The input UVCO of the chip isn't smart enough to protect you from unbalanced unprotected cells. It will drain unprotected cells (if one is weaker than the other) and draw more current trying harder.
Probably up to 2.5 amps. \