Membrane characterization and modeling of membrane processes are essential steps in the development and implementation of new membrane filtration processes. The generalized Maxwell–Stefan equation is frequently used to describe these processes. However, predictive modeling on the basis of characterization experiments using single solutes is still troublesome in a lot of cases. Consequently, a better understanding of the effect of the interaction between different components on the membrane separation characteristics is required. In this work, four well-known commercially available membranes, Desal 5DK, Desal 5DL, Desal G5, NTR-7450, and a newly introduced membrane NF have been characterized. The pore radii of these membranes determined from glucose retention experiments increase in the following sequence: Desal 5DK ≈ NF < Desal 5DL < Desal G5 < NTR-7450. The pore radii and effective membrane thickness determined on the basis of glycerin experiments are within 6% of those determined using glucose. The presence of salt ions, especially of those for which the membranes show low retention, leads to reduction of the retention of neutral components (glucose). The retention reduction, at maximum 10% (absolute) in this study, depends on the membrane selected. For NF and Desal 5DK a high glucose retention drop coincides with a high concentration of the anion salt (Cl−) ions in the permeate, independent of the cation salt ion used. This phenomenon can be explained by several hypotheses. One of these, supported by generalized Maxwell–Stefan model calculations, is that the presence of a pore size distribution leads to the observed shift in selectivity. In the presence of salt ions in the pores, the flux through small pores is reduced to a larger extent than that in bigger pores. Consequently, the retention for glucose drops and a shift in the apparent pore radius is determined. Regardless of the mechanism that causes the reduction of the glucose retention, it is essential that this effect is incorporated in predictive models for nanofiltration to predict the loss of organic components to the permeate sufficiently accurately, since this loss may affect permeate disposal costs or product yield.