Let me know if I'm wrong here.
Throwing more power at an LED can and generally is detrimental. Heat is the main overriding factor. Unless you can remove the heat from the back of the array (primarily) you're on a hiding to nothing as your light output goes down.
One of our improvements a few year ago was to be able to reduce & manage the heat. On our units we could achieve 25% more light output from only 10% more power - Currently with a modest 160W input we're equivalent to a 2500W tungsten or 700W HMI light 12800cd
In my world Lux is the meaning ful measurement since it measures the light output at a given distance - so for the 160W light above, we'd get 1423 lux at 3mtr, 512 lux at 5mtr and 128 lux at 10 mtr - and that's measurable across the while field angle - we'll zoom from 8 degrees to 55 degrees, with no bright spots Measured in Candela that would be 12800cd
Lumens also aren't a great guide, I'll pretty much guarantee most lights are exaggerated, and it doesn't' take into account light losses down stream - so photonic heating of the collimator and the lense. Put a light meter against two equally advertised units and I'll guarantee you'll have significant variances
So in @stuartv case what output is he expecting at what distance. But he makes a valid point of the output dropping with battery voltage (and as the thermal temps rise)
I would love to put dive lights on our test bench and conduct a proper evaluation it'd be really interesting - unfortunately no time since our "side project" is developing virus killing LED units in conjunction with a number of medical research facilities - which is going well but we're not there yet
Oh and if you want a proper LED - the picture below is our Aurea - it's 1.5m in dia with 2400W input, has a 11 degree spot, and is bloody bright at a bloody long distance (used for tall buildings)
I could probably get it fitted with a goodman handle if you wanted..