Having worked in a bureaucracy, the most likely explanation is that someone accidentally posted a version of the press release that didn't include all the edits from the review process (and then corrected their mistake). It seems hard to believe that NTSB is both nimble enough to revise its press release in response to speculation on Twitter, and at the same time stupid enough to think that the revision would make things better.
It also seems hard to believe that those words would have made it into any version unless the FDR data showed column deflection and/or force. Maybe NTSB decided it should formally eliminate any other possible explanation for those readings before publishing the conclusion that there was actual control column input. Or it just didn't want to feed too much speculation ahead of the report (but accidentally did so anyway).