Thinking about those batteries in the context of a really big LED tail light for a bike leads to wondering about the variation in LED forward voltages; it’s possible to drive LEDs in parallel if they’re well-matched for forward voltage. A quick-and-dirty test is in order to get some first-pass numbers… and I have bags of nominally identical red and amber LEDs.

Applying a fixed voltage that produces 20 mA through 14 randomly chosen LEDs of each color, then measuring the voltage across each diode:

Pushing 20 mA through the five lowest voltage red LEDs requires 9.54 V. Applying that voltage to the five highest red LEDs produces 18.2 mA.

Putting those two strings-of-five in parallel with 9.52 V produces 40 mA total: 16.6 mA in the low string and 19.9 mA in the high string, all measured with a fancy Tek Hall effect probe. No, those aren’t reversed and, yes, I did check twice: it makes no sense at all.

Temperature matters a lot in such measurements and I wasn’t controlling for that, plus I didn’t have a constant-current supply. Better numbers await better instrumentation, but I think binning a couple bags of 100 LEDs based on forward current should be straightforward.