When EXAFS samples are slightly nonuniform in thickness, spikes, referred to as "glitches", are observed in the data. Such features are intrinsic to data collected using crystal monochromators. We have recently developed a model to show why such features, persist in ratioed data. Here we test the predictions for linearly tapered samples. Our results suggest that the glitch amplitude can be significantly reduced by taking a second spectrum with the sample inverted, and then averaging the two spectra.