First off, of COURSE "five hours of A, B, or C" means that you can do any ONE of those for five hours (or a combination adding up to five hours).
Second, "I bet most users will have Wifi on and cellphone on". Umm, yeah, probably, but the cellphone will be in standby, which probably has about the battery life of the Treo in standby, which is around 300 hours (give or take).
Third, "It is easy to manipulate battery life figures - are you testing with a strong cellular signal, or a weak signal? Are these real world, or lab tests?" Of course it is. That's why the Treo's claimed 5 hours of talk time is really more like 2 hours of talk + goof around in WM5 time. All cell phone rated times are in ideal, strong cell tower situations. I expect Apples were, too (I mean, there is no "typical" situation for cell phones, and talk time can go from five hours with a strong tower to about thirty minutes on the fringe outskirts of a tower's range).
So, I trust Apple's "5 hours" estimate just as much as I trust Palms (and Moto's, and Nokia's and Sanyo's, etc). Which is: I know 5 hours is about 2 hours of solid talk/usage time. That's what I get on my Treo, and that ratio is what I got on my LG phone before it. It's not a matter of the tests being rigged: it's a matter of the only common comparison that anyone can trust is "idealized conditions", and everyone (for the most part) degrades from that at the same rate.