A large number of SLA observations at a high along track horizontal resolution are an important ingredient of the data assimilation in the Mediterranean Forecasting System (MFS). Recently, new higher-frequency SLA products have become available, and the atmospheric pressure forcing has been implemented in the numerical model used in the MFS data assimilation system. In a set of numerical experiments, we show that, in order to obtain the most accurate analyses, the ocean model should include the atmospheric pressure forcing and the observations should contain the atmospheric pressure signal. When the model is not forced by the atmospheric pressure, the high-frequency filtering of SLA observations, however, improves the quality of the SLA analyses. It is further shown by comparing the power density spectra of the model fields and observations that the model is able to extract the correct information from noisy observations even without their filtering during the pre-processing.