The semiexperimental (SE) technique, whereby equilibrium rotational constants are derived from experimental ground-state rotational constants and corrections based on an ab initio cubic force field, has the reputation to be one of the most accurate methods to determine an equilibrium structure ( reSE). However, in some cases, it cannot determine accurately the position of the hydrogen. To investigate the origins of this difficulty, the SE structures of several molecules containing either the OH or the CH3 group are determined and compared to their best ab initio counterparts. It appears that an important factor is the accuracy of the geometry used to calculate the force field, in particular when the least-squares system is not well conditioned. In this case, the mixed regression method is often an easy way to circumvent this difficulty.