We perform the same analysis than in *this note* for `exp6` using **1/10-day** model output. The error due to the analysis is reduced as the estimates from snapshots and averages come closer (Fig. 1), compared to previous calculation with model outputs of longer temporal resolution. However, what is corrected is the estimate from snapshots as the error estimated from averages does not change (compare Fig. 1b here with Fig. 4b of *this note*). This suggests that the error estimated from averages is due mainly to the error in the model and cannot be reduced. It is, thus, still puzzling that the error in `exp4`, a simulation that does not satisfy well the PV balance during a cycle, has the same magnitude than the error in `exp6`.