Fully coupled photon-electron-positron computational schemes are needed in a broad range of domains, from radiotherapy and radiosurgery treatment planning to nuclear-reactor safety analysis. Industrial deployment of such schemes remains problematic because of the prohibitively time-consuming charged-particle-transport aspects of both Monte Carlo algorithms and Boltzmann-Fokker-Planck solvers. In this paper, a partially coupled scheme based on reactor-physics models for secondary photons (RPSPs) is presented. RPSPs include thick-target-bremsstrahlung, annihilation, and full atomic cascade models, recovering long-range effects of both electrons and positrons. The RPSP-based scheme is compared with state-of-the-art fully coupled and uncoupled schemes in medical, nuclear, and multidisciplinary contexts. Within tissues, the average discrepancies from the fully coupled scheme are about , with a maximum error below at high energy. The CPU time of the coupled schemes is reduced by 58 and 78 times for 3- and 6-MeV beams, respectively. The discrepancies are also within the accuracy of the cross sections for , , , , , , , , , and slabs. The same discrepancies are about , and , respectively, in the reactor's shielding, vessel and fuel slabs. A power peaking factor reduction of 4.35 times from the state-of-the-art uncoupled scheme is demonstrated for uranium oxide and mixed plutonium-uranium oxide slabs.