A liquid in contact with a textured surface can be found in two states, Wenzel and Cassie. In the Wenzel state the liquid completely wets the corrugations while in the Cassie state the liquid is suspended over the corrugations with air or vapor trapped below. The superhydrophobic properties of the Cassie state are exploited for self-cleaning, drag reduction, drug delivery, etc., while in the Wenzel state most of these properties are lost; it is therefore of great fundamental and technological interest to investigate the kinetics and mechanism of the Cassie-Wenzel transition. Computationally, the Cassie-Wenzel transition is often investigated using enhanced sampling ("rare events") techniques based on the use of collective variables (CVs). The choice of the CVs is a crucial task because it affects the free-energy profile, the estimation of the free-energy barriers, and the evaluation of the mechanism of the process. Here we investigate possible simulation artifacts introduced by common CVs adopted for the study of the Cassie-Wenzel transition: the average particle density in the corrugation of a textured surface and the coarse-grained density field at various levels of coarse graining. We also investigate possible additional artifacts associated with finite size effects. We focus on a pillared surface, a system often used in technological applications. We show that the use of a highly coarse-grained density (a single CV) of the fluid in the interpillar region brings to severe artifacts: errors of hundreds of kT in the difference of free energy between the Cassie and Wenzel states, of tens of kT in the estimate of the free-energy barriers, and erroneous wetting mechanisms. A proper description of the wetting mechanism and its energetics apparently requires a fine discretization of the density field. Concerning the finite-size effects, we have found that the typical systems employed in simulations of the Cassie-Wenzel transition, containing a single pillar within periodic boundary conditions, prevent the complete break of translational symmetry of the liquid-vapor meniscus during the process. Capturing this break of symmetry is crucial for describing the transition state along the wetting process and the early stage of the opposite process, the Wenzel-Cassie transition.