I remember being confused by this too. I presume the reason it works this way is to prevent people from creating tags independent of questions, such that the tag would not be associated with anything.
If you think there's a tag that should exist but doesn't based on a particular question, you can edit the question to add the tag, then edit the "tag wiki" to explain what it's for. You're then free to search around for other questions to which the tag might apply, and add it to them.
Tag maintenance can be pretty tedious because recognizing an issue may mean going through bunches of questions. This is one reason tagging is an ideal topic for the chat room, because you can point something out, say what you've done, and then someone else may decide to pitch in searching and applying.
Another thing to be aware of with tagging are synonyms, which are created by intention to avoid people creating two tags for the same thing, sometimes via typo (e.g.,
raspian). If you find two tags you think are synonyms, please ask about it here or in the chat room.
I wonder if that means it is a usability issue for new users?
A long term trend with interfaces and documentation seems to presume increasing amounts of user literacy. As in, I think someone from 1995 would find contemporary smartphone apps hard to use, notwithstanding the fact they are supposed to be easy for someone who's owned a smartphone for a while.
I'm not at all calling you "user illiterate", Greenonline, just pointing out there is this tension. I've been a daily user of S.E. for more than 3 years and I still regularly get confronted with my ignorance of the interface. Whether this means it is too complex or not friendly enough is obviously a judgement call. There's a lot of discussion of this kind of thing at the overarching meta-site.
A fundamental concept in interface design is ease of learning vs. ease of use. The classic example of this is a GUI vs. a CLI (command line interface). GUIs have a much better ease of learning because you can just look through a bunch of menus to find what you are looking for but aren't sure how to do; with a CLI app you have to consult some documentation. However, GUIs are also slower and more awkward to use if you do know exactly what you want to do and how to do it, hence, they have a worse ease of use.
The S.E. interface probably is geared more toward experienced users, meaning aspects of it sacrifice ease of learning for ease of use. The use (or non-use) of copious "confirm and/or explain" pop-ups is part of that. As a new user, I want all the help I can get. As an experienced user, I want a minimum of steps and clutter.
The best of both worlds could be achieved by providing a choice of styles and some configuration options, but that adds to complexity and maintenance overhead. The actual composing/editing interface here has that in the sense that you can use the buttons at the top or you can just hammer out markdown...