Even if no irrelevant questions are included in the mix, the technique you've described would most likely be scored in the same way as the relevant/irrelevant technique: the polygrapher looks for consistent, specific, and significant reactions to any of the relevant questions. If such reactions are observed to a particular question, then deception is inferred.

Augmenting reactions to a different pair of items during each chart collection would tend to diminish the liklihood that any one question would stand out more than others. However, behavioral countermeasures -- and having an innocuous explanation prepared in advance for any claimed "reactions" -- may well be enough to get one through.

In pre-employment screening, a question about marijuana usage would be a relevant one. It is highly irregular for a relevant question not to be reviewed during the pre-test phase. It is possible that a creative polygrapher in this department is using this question as a "control" stimulus. Alternatively, asking an unreviewed relevant question would be a simple way of rigging a polygraph examination against a particular applicant.