Certain oscillatory features in the primordial scalar power spectrum are known to provide a better fit to the outliers in the cosmic microwave background data near the multipole moments of l= 22 and 40. These features are usually generated by introducing a step in the popular, quadratic potential describing the canonical scalar field. Such a model will be ruled out, if the tensors remain undetected at a level corresponding to a tensor-to-scalar ratio of, say, r ~ 0.1. In this work, in addition to the popular quadratic potential, we investigate the effects of the step in a small field model and a tachyon model. With possible applications to future datasets (such as PLANCK) in mind, we evaluate the tensor power spectrum exactly, and include its contribution in our analysis. We compare the models with the WMAP (five as well as seven-year), the QUaD and the ACBAR data. As expected, a step at a particular location and of a suitable magnitude and width is found to improve the fit to the outliers (near $\ell=22$ and 40) in all these cases. We point out that, if the tensors prove to be small (say, r < 0.01), the quadratic potential and the tachyon model will cease to be viable, and more attention will need to be paid to examples such as the small field models.