When modeling discriminatory trade policies -- such as targeted embargoes, selective quotas, targeted export or import subsidies, or preferential trading agreements -- failure to explicitly include assumptions about arbitraging behavior may yield to misleading results. Quadratic programming (QP), Non Linear Programming (NLP), and Vector Sandwich (VS) models implicitly set the rules regarding the possibility of simultaneous exporting and importing. The result is that many analysis using these models may lead to poor results because the models contain implicit limits on arbitraging which may be at variance with the actual policies and/or country behavior. The paper introduces an alternative spatial model. Its main features are that countries are allowed to switch from one side of the market to the other as prices change, and that the researcher is allowed to explicitly incorporate her own assumptions about arbitraging and/or obtain different possible solutions as a function of different policy constraints or different levels of effectiveness in enforcing such constraints. Two numerical examples, one addressing the 1980 US embargo to USSR, the other, constructed, involving preferential trading, show how the results obtained using the proposed model compare with those obtained by applying the most frequently used spatial trade models.