1) Those tests results would be highly correlated, so you wouldn't necessarily be much more than 99% accurate still.

2) If the prior probability of someone committing a crime on a given night is low, you are in the classical example of false positives doing harm via Bayes rule.