Terretta covered it as well as I could have, so I'll just elaborate on a couple things.
Using the word "content" is a lot clearer here: the point is that it's non-structural, rather than that it is non-semantic. It doesn't make sense to hypertext structural elements, but it does make sense to hypertext pieces of content.
Let's say I have a massive corpus. That corpus is composed of text, images, perhaps some audio and video files, whatever. That corpus is structured using markup, by saying "this block is a quote of something else; this block is the header; this block is a list and these blocks are its items" and so on.
That all assumes it's text, of course. If it's a 2d image, that might actually be an image map, in which certain portions are sliced off, defined as subsections (much like an UL tag with LI children), and sensibly linkable. Differently, an audio file might have specific time intervals linked elsewhere; we don't have a UI for dealing with that purely yet (and maybe will never find a reason to have one). And YouTube & co have demonstrated ways of doing video linking. If you link to a specific second of a YouTube video, that's no different from a table of contents in a hardcover book: it's the automatic movement that makes it hypertext.
That's what hypertext is. It links two pieces of content together. An anchor tag is just how we've done it so far.
And I need to emphasize that this is an academic perspective and mostly useful for thinking about how to design a browser's API. And I'm overwhelmingly not qualified to do that. As a web developer, it's mostly impractical to think about this at all: it has nothing to do with developing web apps.
That's not what he said. Also, those are not similar things. A, with or without HREF, can wrap the other two. Images and text are symbols of meaning where the text defines itself as to what that meaning is, and the image defines itself as to what that meaning is.
A blockquote indeed means something, but it doesn't define itself. It just sets off text from other text. It needs text (one of those two things above) inside it to convey the quoted meaning.
Put another way, a blockquote is not content. Text is content, as is an image.
Bold is a style. Hence it being phased out in favor of emphasis (em).
Seriously, your entire argument right now looks like: "there are only three semantic things - anchors, images and text". It ignores the notion of headings, lists, partial references, nested references, sections, etc. These all have textually semantic meanings apart from styling.
Are you just trolling, or are you honestly refusing to acknowledge that there are semantic meanings for things other than the three I mentioned?