Narrowing the search for optimal call-admission policies via a nonlinear stochastic knapsack model