I'll try to explain it again:

If you create a setting where a core part of the setting is that there's all these different races interacting in a rich, vibrant, cultural melting pot, but all your players choose to play humans, then you have a complete mismatch between the setting you created, and the experience the players are having.

This is a problem.

It's not a problem that "players are doing what they want". The problem is that the reality of your game experience is fundamentally different to the setting design you've written. You have a setting document that says one thing, and a playerbase experiencing something different. The disconnect might seem trivial or unimportant to you, or you might not care - but the result is that your setting document is fundamentally inaccurate to the reality of play.

For a designer, this is a problem.

BG3 is a single player RPG where an individual player can make whatever decision they want and experience the game the way they want to play it. I'm not trying to claim this specific problem is an issue in BG3. The only reason I brought that game up was that they publicly released statistical data on millions of players, so it gives good data for the proportionality of player choices.

For most tabletop settings, this isn't (usually) a major issue - a character party is typically on the order of 4-6 players, if they're all humans, that's fine. It's the duty of the DM to make sure that the NPCs and the setting are accurate if that's a thing they care about. It can be a problem if your game is fundamentally about exploring these different perspectives, which some indie-RPGs are focused on.

This is mainly an issue in large-scale social play games, like MMOs and Fest-games, which can easily result in this disparity between setting design and play experience.