In this paper a simple dynamic optimization problem is solved with the help of the recursive saddle point method developed by Marcet and Marimon (1999). According to Marcet and Marimon, their technique should yield a full characterization of the set of solutions for this problem. We show though, that while their method allows us to calculate the true value of the optimization program, not all solutions which it admits are correct. Indeed, some of the policies which it generates as solutions to our problem, are either suboptimal or do not even satisfy feasibility. We identify the reasons underlying this failure and discuss its implications for the numerous existing applications.