This sounds like a bug, which I will look into.  In the meantime, my experience has been that the inclusion of covariates hasn't mattered (though, that depends on the intercorrelation of course), so you are probably ok skipping that...