Liquid crystal variable retarders (LCVRs) are often used in Stokes polarimeters as they allow the measurement of different polarization components by applying an electric field that manipulates the induced retardance. However, the optical retardance introduced by these devices is in general not homogenous across the aperture. Another problem with this type of devices is that the fast-axis orientation is not homogenous, and it changes with the applied voltage. For the optimization of polarimeters, in terms of the noise amplification from the intensity measurements to the polarimetric data, the condition number (CN) is often used, but the effects of LCVR spatial variations are not considered. This paper analyzes the impact of errors in LCVRs in a set of optimized Stokes polarimeters simulated by adding errors in the induced retardance and fast-axis orientation. Then, the CN is calculated to observe the effect of these errors on the optimization. We show how errors in the LCVRs lead to different impacts in the polarimetric measurements for different optimized polarimeters, depending on their experimental parameters. Furthermore, we present the propagation error theory to choose the best experimental parameters to reduce the nonideal effects in optimized polarimeters.