The structure solution step is definitely one of the most difficult problems (along with indexation) if a crystal structure has to be determined from powder diffraction data. The basic reason behind this difficulty is the severe loss of information because of reflection overlap, or, in other words, too many parameters have to be determined from too few observations. Hence, additional information is required in the structure solution process, e.g. molecular structure, information concerning chemically/physically reasonable regions in configurations space, space group etc. The program "Endeavour" is designed to use these data in the solution of crystal structures from powder diffraction patterns. It uses a combined global optimisation of pattern difference and potential energy and has been successfully applied to a large variety of inorganic crystal structures. We found, however, that the inclusion of more and more information may lead to the correct solution becoming an increasingly sharp and narrow minimum on the hypersurface of the cost function. Thus, the solution may become nearly "invisible" for a conventional global optimisation method like simulated annealing. A variety of possible solutions to this problem will be discussed.