You can obviously buy newer tech for even better power efficiency, but that gets into a discussion of accounting for capital cost vs recurring.
I just put together a home server, I need to get a power meter and see what it's using.
Theoretically if I pegged it out with the screen on max brightness I'd get about 90 minutes out the battery (which is about what linux estimated running stress-ng) in reality depending on what I'm doing I get 5-6 hours and if it's light surfing on an evening over 7. Currently it's saying 3 hours 58 minutes @33% charge, I'm actually surprised how long it lasts given I didn't expect it to but looking at powertop it spends a good chunk of it's time running at about half it's top clock rate and it rarely bursts.
Though to be fair I didn't buy this laptop for it's battery life (I can swap those out if needed) or weight (not that it's excessive) but because it sat on a local maxima for what I wanted (fast quad core with HT, relatively light, relatively good battery life).
The GF's little Asus regularly goes over 10 hours with her using it for the usual net stuff and it's got a tiny battery, it's a cracking little machine but Intellij would wreck it.
Clearly that's not a realistic usage, but beyond that I'd say 35W is a high estimate even during use. CodingHorror has some numbers for a Dell XPS M1330 [1], a laptop from 2008 with a Core 2 Duo (2 GHz) and 13.3" display.
Idle at max screen brightness is 20W, wifi bandwidth test bumps it up to 24W. Running an instance of prime95 pushes it up to 50W, but most people don't do that. Especially with hardware accelerated video, I don't think people are spending the majority of their computer use with the CPU pegged.
A home laptop probably spends 18+ hours a day off/asleep, and the maybe 6 hours of use aren't going to all be heavy loads.
[1] https://blog.codinghorror.com/revisiting-how-much-power-does...