The study does not claim unconditional behavior or absence of other factors. Merely a correlation between one attribute and another. So if you build a sufficiently extreme scenario then other factors may eventually overpower that preference. But the correlation would still suggest that the threshold at which one would abandon support for free speech would be different.