Thanks for the clarification.
I agree that in practice today, most of the small amount of leftover space we observe is explained by GBT/coinbase sizing and template construction details. That’s a well-known limitation of the current interface, and not something the post was meant to dispute.
The point I was trying to make is a narrower one: leftover space, by itself, is not sufficient evidence of non-optimal miner behavior, because block assembly is driven by marginal trade-offs. That statement is logical rather than empirical, and remains true regardless of the specific implementation details that dominate in practice today.
If that distinction wasn’t clear, that’s on me.