Per the following report from Light & Motion it looks like a lot of bike LED headlights outputs are grossly exaggerated. The testing was reportedly done by an independent laboratory but there is no way to tell if the Light & Motion lights were cherry picked for the test. The other brands measured results are mostly pretty lousy compared to claimed Lumens outputs.
I get the impression that for most LED lights the listed light output is based on LED manufacturers claimed output levels which presume that the driving circuitry is optimized for ideal LED driving voltage and current levels and the LED chip is kept properly cooled to optimum temperature. As LED manufacturers light outputs are apparently based on LED junction temperatures of 25 degrees C there is no way to normally keep an operating high power LED near that temperature. Also most light manufacturers probably do not correct for lens and reflector losses. Unless anti reflective multicoated a front lens will lose about 6 to 10% due to surface reflections. Depending on the batteries used their charge state also will affect results for battery powered lights.
The only LED headlight makers I know of who clearly claim their Lumens outputs are based on actual testing of production units using an integrating sphere are Niterider and Light & Motion.
Personally I wish there was a law making false or misleading advertising or published specifications illegal for all products. Of course all the LED lights makers would have to do then is make the disclaimer that light output is based on LED manufacturers ratings. I have seen that done too, mostly by flashlight manufacturers.
I would also like to see the makers of Dynamo lights provide a light versus speed curve for their lights with a typical dynamo hub. Some of the high power dynamo lights do not reach full output until quite substantial speeds are reached. Examples are the Supernova E3 Triple and the Light On! Dynolight.
End of Rant!