Just because an interface object behaves in a way reminiscent of a physical object does not mean it is that object. It has an entirely different mechanism. What you're saying is that one kind of mechanism should be patentable while another should not, but it's not clear why there is a distinction.
Read the patent. It describes the specific way the iPhone's unlock screen works, which to my knowledge (and evidently that of the USPTO) was novel at the time it was patented. If you want to know what is supposedly innovative about it, that would be the logical place to look, wouldn't it?
With that in mind, I'll ask again: is theres something specific about that patent that is amiss, or is the whole idea that software-defined interface behaviors can be patented wrong to you?
Is it still (really?) "non-obvious" that GUI elements may mimic physical devices?
Irrelevant. There is no patent on "GUI Element That Mimics Physical Device". That's a concept, not a thing that can be patented. The idea of software interfaces that behave somewhat like physical objects may not be novel, but that tells you nothing about the novelty or lack thereof in any specific interface.
What is so innovative about a virtual toggle switch?
Depends on the switch. But again, this is not a patent on "the idea of virtual toggle switches", it is a patent on the specific way that Apple's unlock screen works. It should be obvious to anyone that there is more than one way to implement an unlock screen, few of which would infringe on Apple's patent, and many of which don't involve ersatz switches at all.
It seems to me that if such things can't be copyrighted, or trademarked, aren't covered by design patents, and shouldn't covered by utility patents, then choosing to implement something in software rather than hardware would be a losing proposition for anyone trying to differentiate or innovate.