The funny thing is there's not that much difference in the raw bulb efficiencies - LPS is slightly lower, but not dramatically. But LPS have more fixture losses (the bulbs are omnidirectional), ballast losses, inefficient dimming (and only to slight degrees) for morning/evening, and so on. LED's high CRI (vs. LPS's virtually nonexistent CRI) can sometimes mean you can use less light, too. In practice, you can use something like 30-40% the electricity with LED streetlamps as with LPS, despite the raw bulb figures not being very different.
But no question that LED is much worse for astronomy. LPS was almost a pure spectral line, very easy to exclude. White LEDs use phosphors and deliver a broad spectrum white (apart from the source's blue peak). Pretty much the worst spectrum you could have apart from incandescent. : It kind of sucks that we have to use phosphors in the first place, as they're wasteful - but unfortunately, LEDs have a "green gap"; in the green spectrum, existing tech has a dramatic drop in output and efficiency (and corresponding increase in price) - while green is what our eyes see best. There's still ongoing work to fill it, but it's been a slog. If we could skip the phosphor and just have a mix of different coloured LEDs, the fixtures would be even more efficient, and it would be much easier for astronomers to filter (esp. if the frequencies were standardized)
