Your question/remark immediately reminds me of the work of Stephen Toulmin, The Uses of Argument, and later works. Toulmin deals with the issue to what extend (argumentative) rationality is universal or field-dependent. His answer is that indeed it is field dependent, but that in a procedural sense we can propose universal structures. This has been a major insight underlying all later argument theories that in fact also try to answer your question.
References in Chapter 6 of my book: gold mining (available on this site)
By zermelos theorem, even Grand Master chess players cannot be perfectly rational, since 100% of chess players have lost chess when playing white, and when playing black.
Rationality can be considered structurally bounded. In a bounded rationality model, perfectly rational behavior applies in the limit where the probability that a decisionmaker make utility-maximizing decisions approaches 1. See, for example.