Abstract \textcolor{red}{In the standard Bell experiment, two parties perform local projective measurements on a shared pair of entangled qubits to generate nonlocal correlations. However, these measurements completely destroy the entanglement, rendering the post-measurement state unable for subsequent use. For a long time, it was believed that only unsharp measurements can be used to share quantum correlations. Remarkably, recent research has shown that classical randomness assisted projective measurements are sufficient for sharing nonlocality} [Phys. Rev. Lett. 129, 230402 (2022)]. Here, by stochastically combining no more than two different projective measurement strategies, we report an \textcolor{red}{experimental} observation of \textcolor{red}{double Clauser-Horne-Shimony-Holt (CHSH) inequality violations with two measurements in a sequence made on each pair of maximally and partially entangled} polarization photons. Our results reveal that the double violation achieved by partially entangled states can be 11 standard deviations larger than that achieved by maximally entangled ones. Our scheme eliminates the requirement for entanglement assistance in previous unsharp-measurement-based sharing schemes, making it experimentally easier. Our work provides possibilities for sharing other types of quantum correlations in various physical systems with projective measurements.