That is surprising. I would expect them to be about the same. If anything, it seems like more mass and heatsinking might draw more current because the heat wouldn't just build up but would be moving on. A hot LED isn't as bright, so maybe it is also drawing less power?
this might be the cause of the resistance of the switch too, i rmb a thread about different type of solarforce tailcap switch being testing and some of the tailcap switches have a bigger resistance than the others and thus result in lower current draw.
I think its fundamental torch design, the weight is almost irrelevant enough to be ignored. Its more likely the contact quality, LED variation, driver variation?
Are you changing the driver only, or is the emitter and the driver combination the same in each different torch.
It's not really the mass but how that mass is used. A chunk of lead bolted to the tailcap (or even the emitter) will do little good - the thermal conductivity of lead isn't that good. A heavy chunk of copper or aluminium - preferably with the LED directly attached to it (Which requires a reflow oven) - will do very well. See here for how it ought to be done.
All emitters can very in their voltage drop. Variation is increased between different output bins, but even within the same output bin, there is still variation. Vf binning isnt common practice yet.
Think of it more like,
Tint bins vs output bins, their somehow related, but in essence their different entities, and can affect the outcome in its own way. In the same way, output isn't always related to the forward voltage of the LED. So regardless of what LED output bin you have, your likely to see small differences in LED forward voltage, and therefore differences in the power drawn by the driver board.
There could also be lots of other things, like the wires, lengths, soldering quality, contact quality between the DMM and the torch, even the way the driver responds to the LED and the DMM.
Also the contacts of the driver to pill also are worth a mention!
My opinion mass has nothing to do. For long runtime similar mass bodies can be distinguished also by the total square area surface. The ability for a host to transfer heat is surely more affected by surface area than mass (within reason). Look at CPU heatsinks. Check the surface area and mass. Stick a led on it and wonder why your 3 times more heavy flashlight barely cope with the emitter of choice and a skinny passive cpu heatsink works a treat.
A Intel i7 stock heatsink (nothing special there, not even with heatpipes) can run a XM-L decently driven indefinitely in passive mode if there is at least some miserable air moving around.
Couldn't it have something to do with the diameter of the leads. I have a C9 drawing only 2,2 A from a 2,8 A driver. I suspect the leads are to thin there. The original driver it came with was only 1,5 A before I replaced it, but I didn't change the leads.