Large datasets produce large outliers. People who are tasked with dealing with these outliers can sometimes fool themselves into thinking that their small sampling represents something more systemic.
It’s always fun to call people naive and there’s hardly a good comeback for that. But that doesn’t make it true.
Obviously we strive for 100%. Why not? But don’t make believe there is a systemic issue here when there isn’t. Beware of confirmation bias, too.