Things like hair and bite mark matches for criminal cases should have to done using "blind" methods with samples from multiple different person's presented and the scientist unaware of which belongs to the suspect. That would require a level of objective certainty that couldn't be so easily and perhaps even unconsciously inflated.