That sounds right. They'd be rendering the frames into buffers somewhere that aren't mapped to a physical screen, so they wouldn't be bound by the properties of the screen.
It's possible that it's also a way to get numbers above 60fps, in case a particular phone prevents screen updates faster than that.
And in such a case, the benchmark then is simply measuring a single aspect of iron and is not measuring at all the end-to-end performance as experienced by the user.