The flaw of game theory is the same as that of contemporary macroeconomics. It is useful in that it provides a mathematical interpretation of the things you are talking about, but its application to real-world phenomena suffers from its dependence on utility functions that have too many degrees of freedom to realistically nail down through observations. No matter the amount of data that goes into the construction of your utility function, the same data could be explained equally well by infinitely many other such functions. Beyond the prediction of simple games - hence the name - game theory becomes too underdetermined to provide models whose predictions could be scientifically validated.