The way I look at it this kind of phenomenon is like Liar's paradox; as soon as someone sets forth a system describing what constitutes rational behavior, it's human nature to think about that system and relate to it in various ways, which has the potential to generate preferences which aren't comprehended within the system (due to their ability to refer to it [explicitly countercultural preferences]). However since most systems of rationality are quantitative the system could remain valid as an approximation.