The -unequal_variance option is not allowed when a covariate is present in a model with two groups of subjects, but that does not look like your situation. I'd like to find out what's going on, but unfortunately I'm currently out of office. Are you happy with results without the option right now?