Not a polygraph examiner, but I was trained on the theory in conjunction with the PCASS, a sort of lay polygraph we used on Afghans, Iraqis, and TCNs in the Army. As a side note, I'm pretty convinced the PCASS was not any more reliable than chance, but it did give an interesting inject vector for other intelligence. Much of the whole thing was about security theater, which is true for the poly as well. Of course, not everyone involved in security theater knows they are playing a role.
Polygraph examiners who take polygraph tests do so in a "directed lie" format. As a comment above hinted at; polygraphs compared a probable lie with a relevant question. In the case of saying that you have never used an illegal substance, that was probably structured as a control question with the relevant question being something like "have you used illegal drugs in the last year." A 100% candid person cannot "pass" a standard poly, and must use a directed lie.
Jon