Lie To Me Paradox

Well, Dr. Lightman's theory isn't a complete hoax, even if its accurateness and applicability is exaggerated very much in the show. Yet, still this doesn't make the TV series "a paradox in itself", since Tim Roth doesn't actually judge the acted emotions of his fellow actors, but just plays a role himself, too. It's an entirely scripted process, so you could call it fictional or unrealistic, but not a paradox really. And as you seem to realize yourself, it's just a TV-show, which naturally exaggerates and dramatizes its premises and techniques to make for a compelling viewing. In the same way computer experts would likely shake their heads over stuff like CSI, actual psychologist are unlikely to be able to work as accurate human lie detectors in the same way Lightman does.

That being said, Cal Lightman and his theories are actually based on Paul Ekman and his research on facial expressions and emotions, who also worked as a scientific consultant for the show. Lightman's concept of microexpressions is also not entirely fictional. The most generally known contribution of Ekman is his work on the Facial Action Coding System (FACS) for classifying movements of facial muscles and deducing expressed emotions therefrom. However, from my limited experience and exchange with actual psychologists employing FACS, I can tell you that those things hardly work as accurately and reliably as depicted in the show, especially not at the detail level shown there. And of course the actors shown in the TV series do those expressions deliberately (and likely exaggeratedly) as part of their acting. But this is part of the shows dramatization.

So no, the show is not "a paradox in itself". And while the depicted theories and methods are far from working as reliably in reality as shown, they're based on actual psychological research.