For a flashlight, the total system efficiency is what matters. That is the emitter, driver, optics/reflector, lenses, wires and contact resistance and the effect of heat on those.

That’s why it is useful to measure the output over time in lumen hours for the complete runtime of the light. Knowing the total battery capacity in watt hours adjusted for different runtimes/current, it’s simple to calculate the lumens per watt which takes into account everything. Or at least approximates it very closely as long as your numbers are accurate.

But since it is rare that a light is used for prolonged periods from full battery to empty, it’s more pertinent to look at the system efficacy at lower levels.

For bare driver tests, HKJ did them in the past: Index of led driver tests