The Marshmallow Test -- an assessment of short-term vs. long-term trade-offs -- is one such case. It seems that the outcomes are far more reasonably explained by whether an individual has been exposed to a high-trust or low-trust environment, one in which promises are kept or broken. That is, even under controlled conditions, the cognitive priors the subjects bring to the experiment differ widely, and interpretations of claims differ.
Sunk cost strikes me as similar.
In the case of loss-aversion, there's the issue that under different circumstances, a given loss may represent a minor setback, an entirely inconsequential event, or a major, life-changing precipice. The distinctions are highly contextual, and depend on both personal background and circumstances.
This applies, incidentally, to organisations and firms as well as people. If you're flying along with ample cashflow, a $5 billion penalty (a magnitude recently experienced by a large tech firm) could be tolerable. If you're scating on a wing and a prayer, tight margins, and market perceptions subject to wild swings (say, as a short-term office-space "tech" startup headed by a dynamic leader with flexible moral and epistemic standards), a few well-timed blog posts might prove disruptive if not fatal.
Context matters. Discontinuities exist. Priors differ.