That said: even if they were accurate, we'd still need to examine how may signature match failures result in detecting actual fraudulent voting, and how many signature match failures don't (ie they are false positives that serve to disenfranchise voters).
It turns out that the vast majority (97%) of ballots rejected due to signature match issues are likely valid. IOW: we'd disenfranchise 32 people for every case of fraud occurring. I have personal doubts that the rate for fraud was even that high (3% of the % of ballots rejected for sig mismatch), as you'd expect some type of criminal charges resulting from it.
Even assuming the 3% of sig rejects number, because that's already a % of sig rejects, it comes out to 3% of 0.2% (the actual number of sig mismatch rejections). This is 0.006% of the total vote.
Additionally, signature match checking disproportionately rejects votes from specific groups, while accepting votes from other groups. In other words, having a signature check (or making it "stricter") allows tuning which demographics you accept votes from. Allowing this type of targeted disenfranchisement is not good