Interesting.. have you got any links clarifying this?-robm
it can be logically explained, no need for a test result, I'll give you how to come to the same hypothesis and the results will back it up.
On an Incandescent configurations you loose light in 3 ways:
1. Reflector surface reflectivity.
2. Losses through the lens.
3. Light out the hole that the bulb came up through into the reflector.
On an LED light #3 is not applicable because the emitter only emits light on a "half sphere" so to speak, instead of all directions, so there is no loss of light "out the back" of the assembly. In incandescent flashlights the loss out the back depends on how far forwards the filament is in the reflector compared to how wide the hole is at the base of the reflector that the bulb is inserted through.
If you want to PROVE this to yourself, go grab a typical everyday maglight with incan bulb. Focus it back and forth while pointed at the ceiling and observe the variations in brightness that occur in the room. Now pick out what looks like the brightest points of focus and dimmest and after doing so turn off the light and observe the position of the bulb. Then imagine the size of the "Cone" of light being lost out the back at that particular focus point and the brighter position will suddenly make a LOT of sense...
On a focused for throw assembly, I estimate that at around 10-20% of that "35%" loss figure is a result of this effect, then 10-20% loss from the reflector emission efficiency and a about 1-5% for the lens.
I think 0.8 is a reasonable conversion for most reflector focused LEDs, and as high as 0.85 for some configuration with special lenses.