This speculation was based upon four reviews of the book. Those reviews were written by Michiko Kakutani, a Pulitzer prize winning literary critic, Lorrie Moore, a short story writer, Charles Taylor, one of the nation's leading film critics, and Brent Staples, one of the nation's leading writers on matters concerning race.
I support Wiki's original decision. First, Roth is infamously known in the literary world to be a grade A prick, and his writing sucks. This man thinks phrases like "fucked the lesbian out of me" constitute literature (read his book The Humbling, no, actually, don't - it's time you can never get back). Second, the four reviewers mentioned above, from fairly different backgrounds, made the same speculation in print. Third, the "secondary sources" only policy is sound for Wiki, as Wiki is basically an encyclopedia derived from gathering accessible secondary source materials.
Cognitively sound readers could have followed the links to the four reviews in the prior version of that Wiki page and determined for themselves that the speculation was speculation, as the text of the Wiki page suggested. That the author denies the speculation is true is not a guarantee that it isn't. And this matter could have been easily resolved by Roth without the New Yorker piece. He could have had one of the many writers he knows write some brief statement, virtually anywhere, stating that Roth had told him or her that the speculation in those reviews was false. That secondary source could then have been added to the Wiki page with content which countered the text that covered the speculative thesis. This would have been easier, and likely faster, than what Roth did to "correct" the page.
I'm sure there are many instances out there where individuals who have Wiki pages concerning them and/or their work would like to be able to contact Wiki and assert "that bit isn't true" and have it removed from the page. But encyclopedias, whether traditional print or new media, can't effectively operate in that manner.