Why is it so easy to underestimate systematic errors when measuring G?

Why is it so easy to underestimate systematic errors when measuring G?