The solar irradiance is the main input parameter when designing solar energy conversion systems. A poor accuracy of the solar irradiance simulation models negatively affect the output energy and the durability of the solar energy conversion system. In the paper, the measured values of the direct solar irradiance in the entire month of July 2016 are analysed and, based on the daily received direct solar energy and the variability of the direct solar irradiance, the days are classified in four categories: clear sky days, partially clear sky days, partially cloudy days and cloudy days. Based on this classification, only four clear sky days were identified in July 2016. The same procedure was applied for the months of July 2013, 2014 and 2015 resulting 13 clear sky days in the entire monitoring period of four years (2013-2016). The measured values of the direct solar irradiance in these 13 selected clear sky days are comparatively analysed against the direct solar irradiance simulated with Meliss clear sky model. Further on, a statistical analysis is performed for the time interval 8:00-16:00 to evaluate absolute, relative and root mean square errors between the measured and simulated values. The results show that the simulation model overestimates, in eleven out of the thirteen clear sky days, the solar direct irradiance in the central part of the day. The measurements were performed in the Renewable Energy Systems and Recycling (RESREC) Research Centre located in the R&D Institute of the Transilvania University of Brasov, Romania.