One of the reasons I wanted to use ARIMA instead of Mixed is because I have to address seasonality, possibly with differencing. Also with repeated measures, aren't there usually many many time series, not just two?