Estimation of the return level of daily rainfall commonly employs the "annual maximum series" (AMS) method. However, the sample size of annual maxima is usually very small. Moreover, if the sample includes extremely large annual maxima values, then these few extreme data affect the accurate estimation of the return level and return period. The meta-statistical approach proposed in recent years enables more robust estimates than the AMS method because estimation is done with a large sample. For this study, we evaluated the accuracy of quantile estimation of observed rainfall data and rainfall data simulated using the Compound Poisson Model by application of the Meta statistical Extreme Value (MEV) distribution and the Simplified Meta statistical Extreme Value (SMEV) distribution. Additionally, we estimated the relation between the fitness of the probability distribution to daily rainfall and the accuracy of quantile estimation by adopting multiple daily rainfall probability distributions. Results indicated that the different sample sizes of daily rainfall in both distributions affect the accuracy of quantile estimates at high non-exceedance probabilities because of the sample size of daily rainfall used for parameter estimation. That sample size affects the number of extreme value samples. Furthermore, the relation between the sample dispersion, depending on the sample size of daily rainfall and the goodness of fit of daily probability distributions, affected the quantile estimation accuracy.