This sounds like apologia from a big closed publisher (AAAS) explaining why open-access is supposedly bad. See, sometimes open-access journals fold, and when that happens, nobody knows what happens to the articles. But they'd like you to ignore two inconvenient facts: First, that traditional closed publishers effectively lose all articles by default by this metric! And second, that the Internet Archive, itself open-access, was essential to conducting the study in the first place!