When investigators got wind of Christopher Steele’s notorious dossier, which made Mr. Page a pivotal figure in a “well-developed conspiracy of cooperation” between Mr. Trump and the Kremlin, it would have seemed like confirmation of what they already suspected.

Having adopted this theory, investigators began to exhibit classic signs of confirmation bias, readily absorbing new information that fit the model they’d built, while overlooking or explaining away facts that didn’t fit. The worst misrepresentations to the court that Mr. Horowitz uncovered are sins of omission — new information the bureau obtained as the investigation progressed that should have led it to question previous representations it had made to the court.

The many layers of review FISA applications go through — laid out in a set of rules known as the Woods Procedures — were ill equipped to detect this sort of problem, because the Woods Procedures focus on confirming that facts in the application match documents in the F.B.I.’s case file. But you can’t fact check a claim that doesn’t exist — which means the process is bad at detecting important information that has been left out.