Actually, that's one of the points that gets confused. It's not the size or wattage of the individual bulbs, it's their total output as a group. 5 watt LEDs don't penetrate any differently than 3 watt LEDs or even big COBs. It's purely total power. The rest of it all has to do with the focus. If you beam the entire panel straight down, all the output goes straight down - better penetration. But outside the footprint, everything else gets screwed.
My numbers are by no means exact, but they're accurate, if I can make the distinction. I think you have to include the ballasts/drivers in the heat equation. As you say, that's where a lot of the heat is produced. The best LED drivers are 95% efficient. Old magnetic ballasts aren't even remotely close to that - an expensive electronic hps ballast runs much cooler.
But I believe these numbers are PAR umols, so they're comparable from source to source. As far as spectrum tinkering, I still think that's small stuff. Higher PAR accomplishes the same thing, unless you already have most of your other environmental variables nailed down and you're looking for those last few advantages.
I remember a video where they were saying that commercial growers were adding in LEDs in between their HPS lights. The LEDs were responsible for completing the spectrum and adding increased levels of terpenes to their grow. Might be BS but I have always grown with HPS only in flower and I do believe I have an increased terpene content since adding the LEDs.
There are lots of other effects that light of different spectra have on cannabis. I am going to purchase some far red leds to turn on right after the garden lights go off. This can help convert phytochrome from Pfr to Pr conformation, a process that normally can take two hours. In effect it could allow me to run 12.5 hours to 13 hours of light and have the plant behave as if it were getting 13 hours or so of darkness. We will see if that is true!