Light Output Measurements
Light output can measured in a few different ways.
Lumens are a measure of the total light emitted by a LED or flashlight (see Wikipedia article for more information on lumens). For scale, a 60-watt incandescent lightbulb produces 890 lumens, a 40-watt bulb is 460 lumens, and a pretty decent flashlight produces about 200 lumens. Lumens are sometimes measured "at the emitter" or "out the front." Emitter lumens will always be higher than out the front lumens because there are some losses involved as the light bounces off the reflector and is filtered somewhat by the flashlight lens.
Emitter lumens LED manufacturers publish specifications listing the lumen output of their LEDs (see Brightness Bins) and sometimes flashlight manufacturers will just take the highest value for the LED in their flashlight and say that is the output. But actual output depends on the voltage and current delivered to the LED. Even then, these numbers are always going to be higher than the actual light output by the flashlight.
Out the front lumens To measure the actual light output of a flashlight, an integrating sphere (see Wikipedia article) is needed that captures all of the light and distributes the light equally so that it can be measured accurately regardless of whether the flashlight has a small hotspot or a wide flood. CPF users MrGman and bigchelis have calibrated integrating spheres, test flashlights that people send to them, and publish their results in posts that are stickied in the CPF LED Flashlights forum. This way different flashlights can be compared using consistent test procedures.
However, a flashlight rarely has consistent output. Instead the output usually drops off early as the LED heats up. So the initial output can be fairly high and then drop by 10-20% within a few minutes. Depending on how well regulated the light is, the output can continue to vary with battery voltage.
FL-1 lumens or ANSI lumens The ANSI/NEMA FL-1 standard requires an integrating sphere and uses a reading taken after the light has been on full power for 30-120 seconds. This number is even lower than "out the front" lumens.
As an example of how lumens can vary, take the 4Sevens Quark MiNi 123. This flashlight features a Cree XP-G LED with an R5 brightness bin that gives 139 lumens at 350mA. The MiNi probably drives the LED at closer to 1000mA, so the XP-G specs indicate 250% of the nominal luminous flux or 347.5 lumens at the emitter. However, 4Sevens advertises 189 out-the-front lumens. When bigchelis tested the MiNi 123 with a CR123A primary cell, he got a reading of 142 lumens when the light was turned on and 137 lumens at 3 minutes (with a higher voltage, faster draining IMR cell, MrGman measured 200 lumens at turn-on and 190 lumens at 2 minutes). Later on, 4Sevens introduced a version of this light with a S2 bin XP-G LED which should be 7% brighter. However for this light they used FL-1 testing and advertise only 135 lumens.
While lumens measure the total output of a light, lux measures the amount of light that strikes a point on a wall. 1 lux is equal to 1 lumen per square meter. This can be measured with a light meter. The problem is that the value will go down as the light is moved further from the wall. Also the value will vary depending on whether the reading is taken in the hotspot or spill of the flashlight beam. Still, it is an easy measurement to make and many flashlight reviewers will use it. It can also be used pretty accurately to get a profile of how the brightness of a light varies with time. Reviewers can take an initial reading, call that 100%, and then chart how the light intensity varies until the battery is drained. See Wikipedia article for more information on lux. 0.25 lux is said to be roughly equal to the light of a full moon. On Mythbusters, they determined you needed about 4 lux to navigate a room if your eyes aren't adjusted to the darkness. They also said that 300 lux is needed to read by.
Candela (cd) is another SI unit, but is calculated in such a way so that it does not vary with distance. This is done by simply taking the value in lux and multiplying it by the square of the distance to the emitter (in meters). Taken at a distance of 1 meter, a reading of 1 lux indicates 1 candela. At 10 meters, a measurement of 1 lux would indicated 1 x 10² = 100 candela. Like a lux reading, it will vary depending on whether the reading is taken in the hotspot or spill of the light, so the measurement is usually taken at the brightest spot. Therefore the candela value is really a measurement of how far a flashlight will throw a beam. One candela represents the light intensity of a single candle and is roughly equal to the now obsolete non-SI unit "candlepower." See Wikipedia article for more information on candela.
An actual distance for throw can be directly calculated from the candela value and will be given in meters. The FL-1 standard does just this by providing a light intensity reading in candela and converting this to throw in meters by calculated the distance at which the light will generate 0.25 lux. Throw is just the square root of the light intensity in candela divided by 0.25 lux which gives meters.