White 5mm LED performance at extreme drive currents

Assuming that you can hold the die temperature constant, as you increase current to a typical 5mm white LED the lm/mA drops beyond around 5-10mA. Has anyone done the experiments or seen published data that shows whether the output flux levels out to a constant value at some drive level or actually drops beyond some current level? The way to conduct this experiment would be to use a very low duty cycle pulsed drive to avoid bulk average temperature rise and short enough pulses that the die does not heat significantly during the ON time of any one pulse. At the other extreme, if the pulses are too short the data becomes tainted by effects of the charging current of the junction capacitance. The photo detector would need to either be very fast if you want to look at a single pulse or the overall photo detector system averages the combined pulse train in a linear fashion. I actually have everything necessary at hand to do this experiment but if it has already been done I see no merit in duplicating the effort. Anyone have any links to info on this question? TIA

The only problem with this is the assumption ...hold the die temperature constant... - which is rather difficult to do in polycarbonate encased 5mm LEDs. Polycarbonate doesn't transmit heat well, and that can contribute to thermal runaway.

Surely the heat is extracted down the leads with a 5mm LED, so you just need to heatsink one or both of those (I can't imagine you'd get that great a thermal solution for higher drive levels though).

Interesting question but how does this relate, practically speaking, to everyday led flashlights?
Have you looked at the data pertaining to the luxeon flash leds?

A bump for this one in case anyone with answers missed it the first time around.

Yeah there were some threads with these measurements in the electronics section a year or so ago.

What about having the LED sitting in a testtube of liquid nitrogen?