What...?

Virtually every type of fraud is first detected by detection of a nominally legal but abnormal behavior, then it's investigated to figure out whether fraud actually occurred. That would – obviously – be exactly how any voter fraud detection scheme works, but I guess you're saying that because the initially detected abnormality is not itself illegal, it wouldn't be investigated?

This is like saying "it's not illegal for all the numbers on your tax return to end in $xxxx.00 and $xxxx.50, so therefore tax fraud is undetectable by means of analyzing numerical patterns."