A common problem in comparative bibliometric studies at the meso and micro level is the differentiation and specialisation of research profiles of the objects of analysis at lower levels of aggregation. Already the institutional level requires the application of more sophisticated techniques than customary in evaluation of national research performance. In this study institutional profile clusters are used to examine which level of the hierarchical subject-classification should preferably be used to build subject-normalised citation indicators. It is shown that a set of properly normalised indicators can serve as a basis of comparative assessment within and even among different clusters, provided that their profiles still overlap and such comparison is thus meaningful. On the basis of 24 selected European universities, a new version of relational charts is presented for the comparative assessment of citation impact.