Transmission fluctuations of log-amplitude and phase can be used to construct coherence functions to infer the statistical characteristics of crustal and mantle small-scale heterogeneities. Recent numerical simulations of wave propagation in random media have revealed an apparent discrepancy between simulations and theory for phase coherence functions. In this paper, we show that the reported discrepancy is a result of the inappropriate phase estimates adopted in some literature. Using unwrapped phase fluctuations, instead of estimates of the fluctuations obtained by picking first arrivals by the threshold method, we find that the theory and the results derived from finite difference simulations are consistent. The method of first arrival picking does not yield adequate estimates of phase fluctuation due to the scattering dispersion of waves in the random medium. Phase unwrapping is viable and allows correct inference of medium properties.