During the last few years, spin-polarized scanning tunneling microscopy and spectroscopy has been developed as a reliable tool to image surface magnetic domain structures of bulk materials as well as thin films and nanostructured systems. In principle, this technique also allows for the determination of the energy-resolved spin polarization of the sample P-S (E) with nanometer resolution, information which might play a crucial role in understanding systems like, for example, non-magnetic adatoms on magnetic surfaces. A main problem in quantifying PS(E), however, arises from the fact that, in contrast to planar junctions, the tip-sample distance generally varies with the magnitude and direction of the surface magnetization, since the distance is controlled indirectly by the tunneling current that is itself spin-polarized. We employ a simple model of the tunneling process to investigate this issue and show that a normalization of the dI/dU spectra with the total conductance I/U is insufficient to correct for their distance dependence.