Phase retrieval algorithm is an effective method to reconstruct the surface distortions for reflector antennas. As the traditional Fourier iterative algorithms usually stagnate at local minima, we previously proposed a global phase retrieval algorithm, named CMAES-HIO, based on the hybridization of hybrid-input-output algorithm and covariance matrix adaptation evolution strategy. We address the problem of selection of the Zernike order used in the phase parameterization for CMAES-HIO algorithm in this paper. By introducing a hybrid evaluation parameter, which combines the algorithm accuracy and time consumption, we utilize the Monte-Carlo method to simulate the algorithm performance under different random surface distortions. Simulation results show that for an unknown surface distortion, a Zernike order of 5 or 6 is probably the optimum for the comprehensive algorithm performance with respect to time and accuracy.