One of the most popular controller design techniques for distributed-parameter systems is the early-lumping approach where an observer-based state feedback is designed on the basis of a finite-dimensional approximation. Its drawback is the appearance of spillover caused by unmodeled states. The approach presented in this contribution aims at reduction of these unwanted effects. For that purpose a class of fictitious outputs is introduced that can be estimated without any spillover. Moreover, these outputs suppress the contributions of the unmodeled dynamics so that the spillover is reduced when the fictitious output is used for the compensator design instead of the original one.