I e-mailed my question to the AMPL google group and now i understand why Minos failed. Could anyone help me how to get Ipopt work? I suspect a change in a tolerance parameter could help (forcing the solver to look for a more precise result). Thanks!