There are two types of study on returns to scale, one is based on the basic concept of economics that inputs change in proportion, and the other is based on the concept that inputs do not change in proportion, and they study directional returns to scale or directional congestion respectively. Based on the fact that the funding scale and team size are disproportionate in the general program of NSFC with congestion, we construct the directional DEA model under congestion and study the directional returns to scale with congestion. In order to avoid choosing directions randomly in the study of the directional DEA model, we fit congestion rates of different directions and we consider the direction of least congestion rate as the relative best investment direction. Then we determine the optimal investment interval. At last, by taking the general program of three disciplines in 2011 as example, we conduct an empirical analysis and the result shows that the research on returns to scale under congestion has more practical value. We also find different disciplines are different between the relative best investment direction and the optimal investment interval to improve performance and they are also different strategies for further enhancing the performance of the general program of NSFC in the future.