It's true that lie-detector test results aren't admissible in court anywhere in the U.S.
That's not because the tests have been proven to be unreliable. Performed by those expert in the discipline, they're mostly accurate. The thing is that "mostly" doesn't cut it for a judicial system that demans absolutes to the extent it's humanly possible to get there. As Linda pointed out several chapters ago, there are documented cases of lie detectors generating false results. The equipment relies on measuring autonomic physical responses generated by people who are knowingly lying for purposes of deception, which won't of course necessarily be so of someone who tells a falsehood.
Nevertheless, behind the legal scenes, attorneys will sometimes place great reliance on such tests in developing courtroom strategies, if the test results are available. And Hugo's findings occupy the same position on the spectrum, with the added advantage that the rigors of lie-detector testing aren't involved. That's why Irving sees him as such an invaluable asset, as his first forays demonstrate. Even a partial success is enough to alter the course of a case, as is seen.