You can either build a dragster light which is unpractical on the higher levels and might even jeopardize battery safety, or you can build a light which can be safely operated on each level and is suitable for users with no knowledge about flashlights.
I like the route sunwayman has taken with these lights.
Ich stimme zu! The lumens produced are limited not only by the LED design (such as forward voltage), but especially by the cell technology. The CR123a Lithium Manganese Dioxide cell has a maximum pulse discharge limit of 3A and a continuous discharge current limit of 1A.
http://lib.store.yahoo.net/lib/botach/nabccr123.pdf
Higher lumens = higher current. lower voltage = higher current. At some point in any attempt to increase lumen output, the current design limit of that cell is reached. A safe design will leave some room for production variations of circuitry and LEDs. Many of the CR123 battery "rapid disassembly" events were with primary CR123a cells!
From a design standpoint: After the forward voltage is subtracted from the CR123a voltage output of 3V, and considering the increasing internal resistance of the cell at currents above the constant rating of 1A, if the driver circuitry attempts to pull more amps to output more lumens, the current quickly exceeds the 3A limit of the design specification of the CR123a cell. Most of the single cell lights being sold today draw well above 2.5A at max output, be they Zebralight AA SC51, or a CR123-based light. The safety margin is becoming rather thin! Worse still, marginal increases in lumens usually result in sharp increases in current with commensurate decreases in run time.
For example: Some current measurements using the 10A range of a DMM (which itself has some resistance) and a Sunwayman V11r:
<note that max lumens means only that the dial has been rotated to the max position. The lumens are NOT the same between the following measurements>
AW Li-Ion 14500 at max lumens = about 1.08A
AW IMR 14500 at max lumens = about 1.35A .
Titanium CR123a at max lumens.= 2.32A
Also, I am able to visually detect a lumen difference in the V11r between 14500 chemistries Li-Ion vs IMR: the IMR has a higher output because of lower cell impedance. Rhetorical question: Is the slight increase in output worth the decrease in runtime?
So the bottom line is that light designers are have to limit the lumens because of the design specs of the battery technology. or change the LED.
The Cree XM-L in various incremental increases in lumen output will not provide the major jump in lumens as found between the XP-G and XM-L. Battery technology changes much more slowly than LED technology. and the matter of heat generated in a single-cell light has been ignored so far. another limiting factor.
IOW, either switch to a multi-cell light, or wait for the next truly new LED design.