At present, the success and tenet of modern adaptive e-learning systems is seen in the efficient semantic enrichment of courseware and in advanced personalization techniques. Automatic adaptation steps, though, lead often to personalized navigational elements in learning assets, which in turn may not fulfill a learner need, because this need changes after delivering the asset. Thus, semantic descriptors should also enable different pragmatic viewpoints in order to solve the conflicts between an assumed didactical recommendation and the sum of possible ?sudden? personal learning needs. Regarding that context, this paper shows experiences gained from the utilization and evaluation of a Dynamic Background Library (DBL) in e-learning systems, and focuses on the usage of Concept Molecules for an enhanced DBL version.