The SRIM simulation code, developed for calculating the stopping and range of ions in matter, was evaluated with the aim of determining its reliability for predicting sputtering yields of solids and ranges of low-energy ions. The simulations were performed using two recent versions, SRIM-2000 and -2003. Irrespective of the choice of the three input parameters of the code (the surface binding, the bulk binding, and the displacement energy), the sputtering yields at a fixed impact energy were found to exhibit an unusual Z1 dependence, which is strongly at variance with experimental results as well as with the predictions of analytical sputtering theory (Z1 and Z2 are the atomic numbers of projectile and target atoms, respectively). As shown in detail for targets of silicon and substantiated for titanium, the ratios of calculated to experimental or analytical yields are generally (much) too large for Z1/Z2≪0.7, decrease rapidly around Z1/Z2=1, and are too small for Z1/Z2≫2 (high-to-low ratios differing by a factor of 3.4 at 1 keV). Additional calculations of the projectile isotope effect in sputtering suggest that the sputtering-yield artifact of SRIM is buried in an incorrect approximation to projectile-target scattering. The low-energy electronic stopping powers of SRIM-2003 were found to be much too low. Differences between detailed and quick calculation modes were also identified. Additional problems are caused by the assumed nonrandom target-atom spacing. The use of SRIM in its present form for calculating sputtering yields and low-energy ranges cannot be recommended. As a by-product of this study it was found by analysis of published experimental data as well as by simulations that, at energies be- low about 5 keV, the projected ranges of heavy ions in light element targets such as silicon increase with increasing projectile mass, a previously unknown effect.