The goal of this work is to highlight and explain the limitations of traditional physical channel models used in network simulators for wireless LANs, with particular reference to VANETs, where these limitations may jeopardize the validity of results, specially for safety applications. The fundamental tradeoff is between simulation time and realism. Indeed, a simulator should provide realistic results as fast as possible, even if several nodes (i.e., hundreds) are considered. Our final goal, beyond this initial contribution, is the development of a stochastic channel model which improves reliability of simulations while increasing computational complexity only marginally. The design of our model is based on the representation of the packet decoding procedure as a Markov Decision (Stochastic) Process (MDP), thus avoiding the computational complexity of the simulation of the entire transmission - propagation - decoding chain bit-by-bit, which can surely provide enough accuracy, but at the price of unacceptable computational (and model) complexity. The paper identifies the key phenomena such as preamble detection, central-frequency misalignment, channel captures, vehicles relative speed, that represent the 'state' of the MDP modeling the transmission chain, and propose an MDP structure to exploit it. The focus is on 802.11p and OFDM-based PHY layers, but the model is extensible to other transmission techniques easily. The design is tailored for implementation in ns-3, albeit the modeling principle is general and suitable for every event-driven simulator.