Many state-of-the-art simulators use nonpredictive phase behavior and property models on surfactant flooding. For example, in the case of phase behavior a variation of the black-oil model is often used in several commercial chemical flood simulators. We examine the various available methods including more recent physical models such as the HLD-NAC equation-of-state and their effects on overall recovery process, front speeds, breakthrough times, and cumulative recovery. We study the change of variables such as salinity by comparing each model's effects on the recovery process. State-of-the-art phase behavior and viscosity models are implemented in chemical flooding simulators to represent several realistic microemulsion flooding cases. HLD-NAC equation of state and Hand's method are used to model phase behavior. A novel predictive viscosity model is also implemented for the first time in the chemical flooding simulator UTCHEM to give a more accurate prediction of viscosity, which affects front speeds and mobility. We study the effects of salinity gradient on recovery in a 1-dimensional homogeneous reservoir. We show that at certain gradients of thermodynamic variables the recovery is hindered by accumulation of water at the trail of the microemulsion slug, because of the composition path entering the Winsor II+ region. Therefore, the driving fluid surpasses the slug and contacts the oil bank directly. This in turn decreases microemulsion phase saturation all the way to immobile values and arrests the slug with small mobilization and recovery of oil. Consequently, capillary number decreases to its original values and residual oil is left out. We show that presence of polymer (to control mobility) - both in the buffer and the chemical slug - has little effect on the appearance of the arrested microemulsion and oil bank, but ultimately placates the negative effects by providing conformance control immediately after the microemulsion phase is trapped. However, when we design the salinity gradient such that the composition path avoids the II+ lobe in its entirety, the slug is robust with or without polymer. A critical slug salinity (HLDL for the case of HLD-NAC equation of state) ensures this condition by eliminating II+ lobes in the composition space. We show that this phenomenon can be prevented using higher surfactant concentrations and/or larger slug size, but it can still occur in the field due to surfactant adsorption.