I do not know enough about K theory myself to be comfortable making the call. Please discuss. If we reach a consensus, I (or Qiaochu) will implement and document the change on the tag-merger thread.
One of the issues is that there is not a lot of traffic in these tags (I imagine). In general I think having more tags is best, this may or may not be one of them. One possible reason for the plain K-theory tag is that there is an arxiv tag that is more applicably described by K-theory than the other two. In the early days, I think people able to make tags were trying to mimic the MO policy of having one for each arxiv area.
I think that there are conceivable questions about K-theory that it might be misleading to tag as just algebraic or topological K-theory. But I strongly doubt those questions would be asked here.
I personally use tags to learn about the views of the asker, and what type of people they want to see the question.
Some of the above are just general comments.