I question whether using glass is a good design decision here. One of the selling points of LED bulbs is that they're unbreakable. As for using 20 XP-Es, they have to be using the bottom of the barrel as far as bin goes but this is understandable given the price point. I'm figuring each LED is being driven at less than 0.4 watts. That should translate into 15% or 20% higher efficiency than driving at the one watt level, which in turn means probably 115 or 125 lm/W in warm white, even using the worst XT-E bins. Actually more because I'm sure junction temperatures are well under 85°C at that drive level. Despite that, the bulb as a whole is getting 84 lm/W, implying optical and driver losses of greater than 25%. That seems kind of high to me. You can design dimmable 120 VAC drivers with about 90% efficiency. That still leaves us well above 100 lm/W. So optical losses are over 15%? I know the goal here is to get something which "looks like a bulb", but most of the time bulbs are hiding in fixtures or behind lamp shades. I think the light distribution would have been similar just using a clear plastic globe, and Cree could have been above 100 lm/W. Don't get me wrong, this looks like a nice product, and I'm sure I'll pick up the 5000K version at Home Depot, but I think they could have done even better. I wouldn't expect them to use anything except the lowest bins given the price, but I think they could have taken the crown for the world's first production 100+ lm/W A19 retrofit with a few minor design changes. I think a clear globe with a removable diffuser might have made more sense. If looking like a bulb is important, you leave the diffuser on. If not, take it off and get 10% to 15% more light.