Partly because it's just the way I'm used to doing it, mostly for ease of access. "Immutable" features (hair & eye color, physical features like moles, etc.) I bind to the activation tag, but everything else such as hairstyle or apparel I don't change.

For my personal use cases I find it good as it just makes prompting quicker and it hasn't negatively affected any of my outputs; hair and eye color can still be modified with respective tags, and I won't bind eye or hair color in characters where it varies among my data such as in characters like Rikka Takanashi or Ruri Gokou.

I personally don't see the point in prompting a character by each tag when there are features that are implicit and can just be simplified down to a singular tag. That said, I'm no prompt engineer and I don't have the time to experiment with training nor look into what popular practices are for better outputs. If there is something I'm overlooking I'm not unwilling to change but I just haven't had a concrete reason to do so