The hung jury in Bill Cosby’s sexual assault trial has puzzled and angered the former comedian’s detractors, since statute-of-limitations laws made this the only possible case to prosecute. But alleged victim Andrea Constand’s sustained, even overeager post-incident pursuit of Cosby’s friendship completely compromised her credibility.

So the prosecution did what sexual-assault advocates often do: they used statistics and studies to argue that women who allege rape should be believed. But here their argument was quite odd. In order to explain why Constand changed her story several times, spoke to Cosby on the phone six dozen times after the incident, and directly lied to the police, they offered expert testimony that lots of rape victims behave like that.

It’s a circular assertion: rape victims often appear non-credible because rape victims often appear non-credible. How do you determine who’s a real victim if you can’t rely on reliability?

If we believe the expert testimony, the obvious conclusion should be to stop relying on rape accusers, period, absent outside evidence. In other words, if we know that victims regularly change their stories, exaggerate details, and minimize the event’s significance, then we’re convicting too many accused rapists – because non-victims do the same things.

I’d rather a system willing to convict with “she said” evidence alone. But advocates for sexual assault can’t have it both ways.

Victims of trauma suffer greatly when not believed, so their advocates work very hard to justify apparently questionable behavior. They cite brain studies that supposedly explain away inconsistencies in the stories of accusers. Participating in a trial itself, as one expert claimed in the Cosby case, can affect memory. If both the first story and the second story sound credible, how can jurors know which one to believe?

Similarly, advocates for rape victims undermine their own credibility when they misrepresent the number of women who were not raped but say they were. The frequently repeated assertion that false claims are rare (less than 10 percent) is grossly misleading. The implication is that the other 90 to 98 percent of rape allegations are by women who were actually raped. Far from it.

The statistics in those studies use definitions like that of the International Association of Chiefs of Police, in which a false accusation must involve “evidence that the assault did not happen” – such as physical evidence or credible witnesses demonstrating the victim lied.

Given the difficulty in proving a negative, the number of rape allegations after a non-rape is certainly much higher than 10 percent. A fair measure of false rape accusations would include all occasions of consensual sex followed by one party claiming it was not consensual. But how do you count that? Under the current definition, even an accusation later withdrawn doesn’t count as “false” absent independent investigation.

Besides, rape is not like pregnant/not pregnant; there are lots of gray areas – one reason rape trials can be so confounding. How drunk is too drunk to consent? How clear does a “no” have to be to count as a no? In one case that got George Will in a lot of trouble, the accuser said she “basically” told an ongoing sex partner no, but did not repeat her refusal during the rest of the encounter. Is a “basically” no a real no? It depends how it was said and how it was heard.

Yet advocacy organizations disseminate wildly exaggerated claims of rapists going free. According to the Rape, Abuse, and Incest National Network (RAINN), only seven out of 1,000 “rapists” are convicted. Since legally men aren’t rapists unless they’re convicted, their statistics technically mean nothing – only that seven out of seven rapists were convicted. Of course, many rapes are neither reported, nor prosecuted, nor end in conviction. How does RAINN decide which of the other 993 cases are real rapes? By believing every allegation?

Any time a rapist goes free his victim can feel victimized again, this time by the society she had trusted would bring her justice. But the purpose of rape trials is not to convict accused rapists; it is to convict rapists. In considering cases like Cosby/Constand, we have two choices. If we believe Constand despite her seeming unreliability by saying rape victims are unreliable, we undermine the entire effort to bring rapists to justice. Alternatively, we can recognize that whether or not an assault occurred, Constand has not been a credible witness, so convicting Cosby in this case would be the wrong result.

David Benkof is a columnist for The Daily Caller. Follow him on Twitter (@DavidBenkof) and Muckrack.com/DavidBenkof, or E-mail him at DavidBenkof@gmail.com.