Subgrid scale (SGS) combustion modeling using flamelet approximations require a model for the conditional dissipation rate. For high Reynolds number flow, statistical independence between the mixture fraction and dissipation is often invoked allowing the conditional dissipation to be expressed in terms of its mean filtered value. This assumption fails for application to pool fires because of the transitionally turbulent nature of this class of flows. In this study, an alternative closure for conditional dissipation rate is proposed based on a transport equation for the mixture fraction filtered probability density function. Application of this model for use in Large Eddy Simulation (LES) of a large, 1 m in diameter, methane-air fire plume results in greatly improved predictions over existing models that invoke the use of statistical independence when comparisons are made to experimental measurements.