To the best of my knowledge, no such statistics have been published. However, a polygrapher initially finding a subject to be deceptive would presumably have a bias that could affect any "re-test."

While it might be helpful to request a different polygrapher for any "re-test," the second polygrapher is also likely to be biased by the outcome of the first "test," including a reluctance to reach an outcome in conflict with that of his/her colleague.

That a polygrapher's expectations can influence the outcome of a polygraph "test" was shown rather spectacularly in a CBS 60 Minutes segment that aired in the 1980s, in which four different polygraphers each found a different person "guilty" of stealing a camera (no camera had in fact been stolen). Each polygrapher was told that a different employee was considered to be the most likely suspect, and in each case, the fingered employee was the employee who "failed."