Well the lumen/watt thing is a bit iffy, even if we accept lumen as an acceptable unit, we're talking some lab setup design to get that 300lumen/watt, probably driving the most efficient current (which is quite a bit less than what we do with LEDs as we very often double almost tripling that current), and what sort of cooling setup are they using? I've seen Intel processors overclocked to 7Ghz, but they were cooled with liquid nitrogen!
Now accepting PAR as metric it should be somewhat easy to check how the efficiency really stands out, of course which bulb you choose on the MH side will make a world of difference too. Simply take measurements (underwater) at a particular depth over an area until the amount of light drops to negligible amounts, then simply come how much power used with a kill-a-watt meter or something and just call it a day. I tried doing something like this the last time I borrowed the club's PAR meter but unfortunately the rocks got in the way and then I just went with the "awww screw it, this is too much work" result
As to the heat myth, I was referring to the need to have chillers and/or fans to cool your tank. Obviously this depends greatly on each setup, how hot is your house, are you using 1200 watts of halides over a 100g tank? etc. I'm fully aware that a metal halide heat source essentially is acting like one of those satellite dish looking electric heaters you see in CostCo, so yeah more energy will be put into the tank. HOWEVER, how much less are your heaters turning on as a result? So going back to the original question that might be fair to take other power that's not used by your tank as a nod towards the halide side of things as far as "efficiency". Now if you need a chiller that runs whenever the lights are on then sure all that goes out the window. But like I said, never used a chiller in my life, and the first decade or so of reef keeping used halides
The reason I mentioned DIY is because inside a canopy (if you have one) there's virtually no difference in terms of quality, and in many cases you can buy an actual better reflector with a MH setup (lumenarc type reflectors). And with a garage full of junk I can throw together a MH fixture without any additional cost, yet I can't throw together a DIY LED fixture for no cost because I'd need to buy all the parts.
There are two knocks against LEDs I see though, first because the equipment is evolving so fast it kind of sucks to pay for equipment and have a better version a year or two later at the same cost, how many versions of the Radion came out? AI Vega/sol/Hydra? Kessil A150/350-160/360/w/e... Second, what's the life span? I mean we all hear numbers of 100000 hours (or whatever) now are those lab guestimations? How much does the light drop? I know I've let MH bulbs go too long because "hey it's making light and looks the same brightness to me!"
. And what happens if it does go faulty? On the hook for another $400 kessil? $800 Radion? did it pay itself off before that happened? Where as the modular nature of a MH makes it bit easier to handle, with LEDs you really are investing in your tank, now LED design has come light years since one of the original PFO Solaris fixtures which seemed to have LEDs dying all the time, but still is something to be mindful of.
Don't get me wrong, I absolutely LOVE LEDs, their ability to be customized and controlled is a very VERY powerful PRO that very often outweighs most cons. My point of arguing now... other than me loving a good debate... is that I don't think metal halides are quite a dead fish yet. That said, after just a trying to do a bit of research before I posted, they do seem to be going the way of the dodo as far as fixtures go, how much longer with the replacement bulbs stick around?