I didn't read the NYTimes article, but the report itself claims to "Include only studies with random-assignment or controlled quasi-experimental designs". They

I didn't read the NYTimes article, but the report itself claims to "Include only studies with random-assignment or controlled quasi-experimental designs". They go on to list some remaining caveats on page xvii, and I can think of one or two, but I would imagine data suffering from the source of bias you describe was one of the first things they set out to discard.