Why is it so difficult to introduce requirements engineering research results into mainstream requirements engineering practice?

While research results in requirements engineering (RE) were developed without much interaction with or impact on practice for quite a long time, Hsia et al. [1] made an honest evaluation of the requirements engineering practice. Since then, some improvements have been achieved, e.g., by some applications of usage scenarios. Still, mainstream practice is to use one's favorite word processor to write down mostly unstructured text in natural language, even though some good experience exists that used RE results in real projects (see, e.g., [2, 3]). In fact, mainstream practice only rudimentarily deals with requirements at all, and what is done might not even deserve to be called ?engineering?.