Someone searching "at the same time" could technically be different times (remember time zones!) for the purposes of the algorithm (is the search during "work hours" or not, etc).
Users checking results on mobile phones compared to desktops... Without more details of how they controlled for these factors, the conclusion doesn't really follow.
Edit: I think to really measure the conclusion, they'd need 87 people in the SAME geo location, perhaps on fresh out of the box devices. That would be the best way to create the "placebo" group for their test, which they don't seem to have done.