No, I'm saying their bar for adding things to the language (i.e. their "significance" threshold, which needs to be defined consistently somehow) seems to be low

No, I'm saying their bar for adding things to the language (i.e. their "significance" threshold, which needs to be defined consistently somehow) seems to be lower in reality than your interpretation—or my default one, for that matter. I'm saying if I use the same threshold they utilize in reality, then yes, the improvement is very much "significant" enough to add.