A thorough understanding of the power consumption behavior of real world wireless devices is of paramount importance to ground energy-efficient protocols and optimizations on realistic and accurate energy models. This paper provides an in-depth experimental investigation of the per-frame energy consumption components in 802.11 Wireless LAN devices. To the best of our knowledge, our measurements are the first to unveil that a substantial fraction of energy consumption, hereafter descriptively named cross-factor, may be ascribed to each individual frame while it crosses the protocol/implementation stack (OS, driver, NIC). Our findings, summarized in a convenient new energy consumption model, contrast traditional models which either neglect or amortize such energy cost component in a fixed baseline cost, and raise the alert that, in some cases, conclusions drawn using traditional energy models may be fallacious.