Most industry studies count all or almost all instances of piracy as substitutions for paid sales. The GAO concluded that each of the studies relied heavily on assumptions about the substitution rate (how much of the infringing activity can be counted as a lost sale), a number that has yet to be independently proven.
Tracking the impact of piracy on paid content sales is hard work. You have to sample enough to draw reasonable conclusions, and you have to refrain from drawing conclusions before you have reasonably good data.
While the ars technica headline (“US government finally admits most piracy estimates are bogus”) seems overly broad, it’s encouraging to read reports that call for data-driven assessments of the true impact of piracy.