I disagree with you here, and but I can agree that this XGD statistic is awful.

I imagine it’s constructed from XG and XGA (which will have different distributions), and not presenting it with these figures really makes it’s value very limited indeed.