Towards Modelling Defeasible Reasoning with Imperfection in Production Rule Systems