It just depends on the 'linux side of things' group your talking about. My T420s runs a full featured GNOME desktop on Arch at 5.5w idle, 7w browsing, 10w loaded. 136 watt hours total battery capacity with three batteries.
"Hours" is a really poor way to gauge these things, because while it may seem practically applicable , we're really comparing apples and oranges.
My T420 before my T420s had a battery run-time of 15 hours. I could fail to mention that it had a slice battery and an ultrabay on top of the primary battery and tell everyone that the particular model and OS ran for 15 hours straight, but that would be disingenuous. If I was a marketer I could run the processor at lower voltages, dim the screen unread-ably, turn WiFi off, etc etc until I got whatever numbers I wanted.
As i'm not a marketer, I also feel as if the claims people make about Apple's hardware being better at power management are also somewhat dishonest.
They are a company that has ultimate say in their hardware platform and software ecosystem. No other computer company locks down production that much in an effort to increase the individual experience of each consumer. That lock-down allows for an amount of control that would be unrealistic to expect from the FLOSS community, as they have little power or say in production and ultimately must code to support a wider range of people rather than a narrow set of known hardware packages that are available.
My point : optimizing narrowly for a single platform is much easier than the opposite; a more fair comparison of battery life would be watts-per-hour usage during a standardized testing/benchmarking suite of some sort, along with the watt hours of battery capacity(although historically handset manufacturers have sidestepped that issue by having the software recognize when it's being benchmarked..). "Hours of use" could be practical if it wasn't arbitrarily inflated or deflated by uncommon usage profile testing and specialty optimizations for the sake of marketing a bigger number.
It's not impossible to get good numbers out of a Linux build, it's just fairly difficult for someone unfamiliar with the system (and sometimes impossible without doing the work yourself with certain hardware stacks). There exists a whole subspecies of Linux users that are somewhat akin to hypermillers who turn this kind of dilemma into a game and compete for lowest wattage.
