Edit: hit HN's comment limit again. Anyway: consider this - what if the test isn't 100% reliable? What makes you think it is - because you want to believe that 100% truth can be established? What if it has a false negative rate of 30%, so 30% of actual rapists (who might already have confessed) would be cleared by the test?
Revealing innocents pressured into plea deals would have social value. Were you under the impression that innocent people never confess?
That said: DNA tests are far more reliable than eyewitness accounts. Innocent people confess to crimes all the time.
You'll easily be spending 100x that much on the trial and imprisonment. Spending an extra 1% to get some hard physical evidence seems like cheap insurance.
I'm sure there are some circumstances where testing the kit is completely redundant. But how often does that happen? If it's rare, why bother carving out an exception?
Additionally, as another commenter mentioned, false confessions exist.
I don't understand your "standard procedure" comment. Either collecting the kits is useful or it is not. If it is, then test them. If it is not, don't collect them. If collecting them is "standard procedure" then change the procedure.
Collecting this evidence with no intent to process it is borderline abusive.
Rape kit collection is one of many reasons why many people do not report their rape to police.