Many methods of optical tissue diagnosis require that measurements be performed with small source-detector separations in a backscatter geometry. Monte Carlo simulations are used to demonstrate that for these situations light transport depends on the exact form of the angular scattering probability distribution, P(theta). Simulations performed with different forms of P(theta) with the same value of ?cos theta? result in the collection of significantly different fractions of the incident photons, particularly when small-numerical-aperture delivery and collection fibers are employed. More photons are collected for the distribution that has a higher probability of scattering events with theta > 125 degrees . For the clinically relevant optical parameters employed here, the differences in light collection are >60%.