New biasing sampling schemes for the "once-more-collided-flux-at-a-point" method for gamma calculations are studied. This technique, designed by Kalos, Steinberg, Kalli, and Cashwell, for neutron point flux estimation cannot be directly applied to photon transport problems. Because of the different interaction processes, it is shown that even if the mean and second moment remain bounded, very large variations of the scoring weights occur for specific collision deviations, which leads again to jumps of mean and variance. Two biasing methods are proposed for the resampled postcollision direction and the next deterministic collision to treat the anisotropic behavior of the coherent scattering that is the main cause of instability. The method is tested on an MCNP4A benchmark. This new treatment of the point flux estimation has been integrated in the TRIPOLI-4 Monte Carlo code. Note that no bias results are achieved with CPU costs, which reserves this method for reference calculations.