I base my 2/3 light transmission on tests done on M@gs on CPF. Stock Mags lose 1/3 of the light. Fenix which use high quality AR coated glass get 81% transmission.
So I do not believe cheap acrylic will give 85%.
You are comparing a sputtered aluminum reflector based light (Fenix) to a total-internal-reflection optic. Total internal reflection is an inherently more efficient process. A very high end glass TIR optic, with appropriate optical coatings can be made significantly over 90% efficiency. One with made with no coatings might be 80-85% efficient. Note these are ballpark, I don't know the specific numbers for the exact parts in question.
An aluminum reflector with similar narrow beam profile* and lumen-gathering capability of a TIR optic will may be 80% efficient with an AR-coated window in front, less wihtout an AR window.
Also, the effect of an AR coating is to block reflections off of the window. 4% of light reflected off the window in a refletor based light does NOT necessarily mean 4% lumen loss -- muhc of the light that reflects off of the glass window will reflect off the reflector and eventually make it back out as spill, though certainly that light will be less useful than light that would have made it as part of the original beam initially.
* In the fenix, most of the light never reaches the reflector surface due to the shallow reflector, and forward-emitting beam profile of the LED. The aluminum reflector coatings used in most flashlights are VERY lossy (evidenced by the fact that stock maglites lose so much, as you pointed out -- in the case of the incan, most of teh light DOES hit the reflector).
In a light like the Fenix, most of it flies straight out the front as spill thus lumen losses are less.
A good TIR optic will have both a reflecting surface, and a refracting surface (lens) to capture light, so that nearly all the emitter lumens are collimated as part of the beam. A reflector producing similr beam profile will need to be very very deep, and will be muhc lossier, all else equal.
I base my 2/3 light transmission on tests done on M@gs on CPF. Stock Mags lose 1/3 of the light.
Stock mags are incandescent lamps, thus they are omnidirectional light sources and and as such have yet another source of lumen loss compared to LEDs: the big hole at the bottom of the reflector.
Also, a bigger fraction of a light from an incan in a mag reflector will actually bounce off the reflector, compared to light from an LED in a mag reflctor. This meanas two things:
1) bighter main beam for the incan, given the same number of emitter/bulb lumens
2) more lumen loss because the aluminum reflector surface is lossy.
IMO this is why the rightful comparison should be between lights with comparable beam angles etc.
One other thing I like to mention.
I bought a
3*Cree bulb a year ago. I compared it to 3*L0D CE strapped together. The brightness level and beam pattern were the same. I turned the lights off and reached to take the LED bulb from the lamp. I almost burnt my hand. The L0D CEs were barely warm. I figured the power supply of the LED must be generating all that extra heat. A very inefficient power supply.
When comparing bulbs it is necessary to compare LED+power supply to CFL tube+ballast to incandescent wattage. The bulb marketing departments tend to forget the power required for the power supply or ballast.
This may not necessarily be because of an inefficient driver. My guess is the L0D's were being moved around in open air, and/or held in your hand.
If the L0D heats up to thep oint where it's warmer than your skin temperature, your hand will actually conduct away a huge amount of heat (your circulating blood will act like active liquid cooling.
I once ran an L0D-CE with LiIon in my hand down in 10 minutes. Barely got warm. I did the same test with one set down a table and nearly burned myself on it. Even larger, better heatsinked flashlights have become extremely hot to the touch when they've come on inside insulated jacket pockets etc.
OTOH that's very interesting that your L0D's didn't get very warm. Are they colored black? I've heard that black heatsink dissipates heat better than bright aluminium or even brighter chromium surfaces. Or maybe there's significantly more surface area in those L0D's than in the replacement bulb. Are you willing to calculate?
black objects will radiate energy much more efficiently. In a place with airflow, convection will probalby be a much more significant means of heat dissipation. As temp goes up, the significance of radiation goes up.
From Wikipedia:
where
Emissivity of aluminum foil is 0.03, anodized aluminum is 0.77
Let's call the aluminum foil emmisivity an even "0"
Let's say total surface area of L0D is 10 square cm.
Let's say temperature of our L0D is 114F, or 45C, or 318k, and ambient temp is 70F, or 20C or 293k)
Additonal Power radiated due to anodized aluminum:
P = ( 0.001 sq-meter ) ( 0.77 )( 5.67e-8 )( 318^4 - 293^4) = 0.125 watts
An L0D CE running off a LiIon battery (would have to be to get that hot) would be dumping about 20x that much as waste heat. So the marginal effect of the anodize won't become significant unless the temperature of the device are hot enough to damage the LED (or the user for that matter)
.
...I also wouldn't mind if I bought an led but with a 70% lumen maintenance over 5 or so years.. Hopefully it would have paid itself off by then.. 
