Most science isn't accurate, that's why it's filled with p-values. They're often only reporting the mere hint of an effect which allows them to distinguish between a control and experimental population.

Polygraphs are almost certainly not reliable enough to convict someone on, but it's not admissible in court. That doesn't mean it can't be used to help coerce a confession or help lead an investigation. And the fact that you won't face up to the extraordinary amount of evidence showing that it can be useful suggests stubborn ignorance on your behalf.