Accurate description of VLE (and VLLE) plays an essential role in many industrial separation processes. Local composition models for the activity coefficient description, such as NRTL and UNIQUAC, have contributed significantly over the past four decades toward the fitting of experimental VLE data, for example, to the design of distillation towers. However, a critical examination of this field reveals that important deficiencies still exist due to the complex nature of the topic. In the present paper, these pitfalls are classified and discussed in detail, and some examples are selected to conveniently illustrate every one of them. We also comment on some aspects that could improve the distribution, evaluation and correlation of these VLE (and VLLE) data that, in turn, could lead to better results and avoid misinterpretations.