Infinitesimal-rotation finite elements allow creating a linear problem that can be exploited to systematically reduce the number of coordinates and obtain efficient solutions for a wide range of applications, including those governed by nonlinear equations. This paper discusses the limitations of conventional infinitesimal-rotation finite elements (FE) in capturing correctly the initial stress-free reference-configuration geometry, and explains the effect of these limitations on the definition of the inertia used in the motion description. An alternative to conventional infinitesimal-rotation finite elements is a new class of elements that allow developing inertia expressions written explicitly in terms of constant coefficients that define accurately the reference-configuration geometry. It is shown that using a geometrically inconsistent (GI) approach that introduces the infinitesimal-rotation coordinates from the outset to replace the interpolation-polynomial coefficients is the main source of the failure to capture correctly the reference-configuration geometry. On the other hand, by using a geometrically consistent (GC) approach that employs the position gradients of the absolute nodal coordinate formulation (ANCF) to define the infinitesimal-rotation coordinates, the reference-configuration geometry can be preserved. Two simple examples of straight and tapered beams are used to demonstrate the basic differences between the two fundamentally different approaches used to introduce the infinitesimal-rotation coordinates. The analysis presented in this study sheds light on the differences between the incremental co-rotational solution procedure, widely used in computational structural mechanics, and the non-incremental floating frame of reference formulation (FFR), widely used in multibody system (MBS) dynamics.