I’ve never thought of Markdown as particularly semantic in nature. Maybe that’s because I compare it to explicitly semantic projects like the Text Encoding Initiative (TEI) and the Data Documentation Initiative (DDI) for social and behavioral sciences.
With TEI, you can markup individual sentences, you can identify text as prose or verse, and you can give meaning to text in many other ways. Now that’s semantic. Markdown doesn’t have anything like that, just lists, paragraphs, italics, etc.
I think a lot of Markdown is more presentational than semantic, so I’m surprised that you have the opposite view. HTML, Markdown’s target, hardly has any semantics. People talk about HTML5 like it added a bunch of semantics, but it’s really trivial stuff like an article tag, a main tag, and so forth. I mean, that’s hardly anything at all. You need to use something like Schema.org markup to add semantics and meaning to HTML.
And I’ve never been sold on the bright line between semantics and style/presentation. I think italics and bold, or what some people insist on calling emphasis, are more than semantic constructs – they’re stylistic too. I think a lot of HTML elements have stylistic functions, and I don’t think there’s a bright line between style and semantics in the English language.