This is a common problem / issue discussed on the forums - see:
https://forum.inaturalist.org/t/overzealous-identification/5975
https://forum.inaturalist.org/t/false-research-grade-observations/14193
https://forum.inaturalist.org/t/gamify-accuracy-award-value-to-quality-not-just-quantity/14428
I think the general consensus is that it’s the least bad system - the current system leans slightly more towards simplicity and inclusion vs. every research grade being accurate. Although the top identifiers sometimes end up having to do quality control in areas they know.