Representativeness of weather inputs is crucial to limit the global uncertainty of building energy simulation results. The length of the multi-year weather data series and the methodology used for the typical month selection largely influence the results of the reference year development process. In this work, we investigate two possible modifications to the EN ISO 15927-4:2005 procedure aimed at improving the representativeness of reference year heating and cooling needs. The first modification maintains the reference years independent of their final use while the second one leads to the development of specific weather files for heating or cooling analyses by introducing weighting coefficients for the different weather parameters. The study is performed for five North Italy localities with 10 or less years in the data-set and for a sample of 48 simplified buildings. Both proposed modifications brought improvements to the representativeness of the reference year results.