Rain gauges and weather radars do not measure rainfall at the same scale; roughly 20 cm for the first and 1 km for the second. This significant scale gap in not taken into account by standard comparison tools (cumulative depth curves, normalized bias, RMSE, …) despite the fact that rainfall is recognized to exhibit extreme variability at all scales. In this paper we suggest to revisit the long lasting debate of the representativeness of point measurement by explicitly modelling small scale rainfall variability with the help of Universal Multifractals. First the downscaling process is validated with the help of 16 rain gauges located within a 1 km2 area on the campus of Bradford University (United Kingdom). Second this downscaling process is used to evaluate the impact of small (i.e.: sub-radar pixel) rainfall variability on the standard indicators. This is done with rainfall data from the Seine-Saint-Denis county (France). Although not explaining all the observed differences, it appears that this impact is significant.