Low-energy ($<$ 300 keV) protons entering the field of view of the XMM-Newton telescope scatter with the X-ray mirror surface and might reach the X-ray detectors on the focal plane. They manifest in the form of a sudden increase in the rates, usually referred to as soft proton flares. By knowing the conversion factor between the soft proton energy and the deposited charge on the detector, it is possible to derive the incoming flux and to study the environment of the Earth magnetosphere at different distances, given the wide and elliptical XMM-Newton orbit. Thanks to detailed Geant4 simulations, we were able to build specific soft proton response matrices for MOS and PN. In this second paper, we present the results of testing these matrices with real data for the first time, while also exploring the seasonal and solar activity effect on the proton environment. The selected spectra are relative to 55 simultaneous MOS and PN observations with flares raised in four different temporal windows: December-January and July-August of 2001-2002 (solar maximum) and 2019-2020 (solar minimum). We selected and extracted the flare mean spectra and count rates in the 2 -- 11.5 keV energy range for the four epochs. After investigating the rate variations among the MOS1, MOS2, and PN instruments, we fit the X-ray spectra using XSPEC and the proton response matrices. The best-fitting parameters derived for the three instruments were compared in order to obtain the systematic errors. There is no seasonal or solar activity effect on the soft proton mean count rates, but we find large discrepancies in the instrument cross-correlations across the 20 years of satellite operations. In 2001-2002, after a few years of operation, the MOS1 and MOS2 rates are similar, and about 20<!PCT!> with regard to the PN ones. After 20 years, PN does not present any variation in its response, while MOS1 suffers a reduction of sim 30<!PCT!>, in addition to the 30<!PCT!> loss due to the damage of two CCDs, and MOS2 is affected by an even worse degradation (70<!PCT!>). The main result of the spectral analysis is that the physical model representative of the proton spectra at the input of the telescope is a power law. However, a second and phenomenological component is necessary to take into account imprecision in the generation of the matrices at softer ($<$5keV) energies. This component contributes for 21<!PCT!> for the MOS and 5<!PCT!> for the PN to the total flux in the 2–5 keV energy range. This study, which is the first application of the soft proton response matrices to real data, shows coherent results between detectors and allows us to estimate systematic uncertainties in the measured spectra of 3<!PCT!> between the two MOS detectors and of 24<!PCT!> between MOS and PN, together with a systematic in the input flux of about a factor of two. They are all likely due to uncertainties in the proton transmission models, with the presence of additional passive material in front of the front-illuminated MOS, and element deposition on its electrode structure across the mission life. Dedicated studies and laboratory measurements are required for improving the accuracy of the proton response files.