The present study compares the ability of three semipredictive approaches, namely, the global phase diagram based semipredictive approach (GPDA), predictive Soave-Redlich-Kwong (PSRK), and linear combination of the Vidal and Michelsen mixing rules (LCVM), to describe the experimental data in the binary homologue series of ethane-n-alkanes. The applicability of GPDA has been extended by attachment of an additional empirical functionality. Now it can predict not only the data of a homologues series to which the reference system belongs but also other series that include solvents from the same chemical group. Yet, both GPDA and PSRK appear as entirely predictive models. However, LCVM has a correlative character rather than predictive. Although the models under consideration yield rather similar results for the light homologues of the series, starting from ethane-n-decane, GPDA shows a clear superiority over PSRK and LCVM. Moreover, it is demonstrated that the G(E) mixing rules of both PSRK and LCVM generate the nonphysical behavior of liquid-liquid equilibrium. This result strongly affects the reliability of these models in predicting the data of heavy homologues. In contrast, GPDA predicts robust results even for very asymmetric systems. These facts confirm the clear advantage of an approach based on consideration of the complete phase diagram and not of its separate parts.