Maybe I am misunderstanding your point, but I believe what you are describing is what is referred to as VR (Virtual Reality), not AR (Augmented Reality).
From what I understand, I believe other commenters in this thread refer to AR as a combination of 'real' reality and computer generated overlays. This is opposed to VR where all of reality is generated. So when we talk about AR headsets we're talking about things akin to Google Glass or Microsoft Hololens if you're familiar with these products. In that sense you are never 'inside' AR (with a headset or phone), it simply augments what you currently experience with extra data. With a true AR headset it is still possible to interact with people in the real world, you can turn on/off the extra data easily (since you are still immersed in the real world) and the computer generated overlays fit in with your real surroundings in a way that is simple and intuitive. Granted such a headset does not yet exist, but I believe that is the end goal.
What you describe as 'stepping inside the Metaverse' with no possibility of interacting with the real world is exactly what VR is all about (full immersion). (As an aside, the Metaverse as described in Snow Crash is exactly this: a Virtual Reality.)
As for the headset vs phone debate, personally I believe that the applications of AR in a non-headset configuration will be limited. There are definitely applications for one-off utilities which don't require constant use (ie. don't require gorilla arm), but having it available at all times and transparently via some sort of headset is where I believe it will really be useful.