The 3D-QSAR technique MaP (Mapping Property distributions of molecular surfaces) characterises biologically active compounds in terms of the distribution of their surface properties (H-bond donor, H-bond acceptor, hydrophilic, weakly hydrophobic, strongly hydrophobic). The MaP descriptor is alignment-independent and yields chemically intuitive models. In this study, the impact of different operational parameters on the interpretability and model quality was investigated. Based on a set of antimalarially active naphthylisoquinoline alkaloids the effect of hydrophobicity assignment as well as the differentiation of H-bond propensity was evaluated according to a full factorial design. It turns out, that including different categories for H-bond donor strength significantly improved interpretability, reduced model complexity, and made possible the derivation of a novel pharmacophore hypothesis for this dataset. Further analysis of the factorial design reveals, that MaP models are robust to parameter changes and generate consistent models for different parameter settings.