Small angle scattering by dust grains causes a significant contribution to the total interstellar extinction for any X-ray instrument with sub-arcminute resolution (Chandra, Swift, XMM–Newton). However, the dust-scattering component is not included in the current absorption models: phabs, TBabs, and TBnew. We simulate a large number of Chandra spectra to explore the bias in the spectral fit and N H measurements obtained without including extinction from dust scattering. We find that without incorporating dust scattering, the measured N H will be too large by a baseline level of 25 per cent. This effect is modulated by the imaging resolution of the telescope, because some amount of unresolved scattered light will be captured within the aperture used to extract point source information. In high-resolution spectroscopy, dust scattering significantly enhances the total extinction optical depth and the shape of the photoelectric absorption edges. We focus in particular on the Fe–L edge at 0.7 keV, showing that the total extinction template fits well to the high-resolution spectrum of three X-ray binaries from the Chandra archive: GX 9+9, XTE J1817-330, and Cyg X-1. In cases where dust is intrinsic to the source, a covering factor based on the angular extent of the dusty material must be applied to the extinction curve, regardless of angular imaging resolution. This approach will be particularly relevant for dust in quasar absorption line systems and might constrain clump sizes in active galactic nuclei.