The point of dunning-kruger effect is not that there's no correlation- it's just not linear. Studies show there is a tier of most competent people who tend to rate themselves lower than the secondmost tier of competence, who rate themselves higher due to a combination of signaling effects and dunning-kruger ("too dumb to know how dumb you are," or more accurately, "not quite smart enough to know that there is a lot you don't know").