The infiltration of stream water in the sediment and its return to the stream—a process known here as hyporheic exchange flows (HEF)—is a critical control of the structure and functions of the stream ecosystem. River restoration programmes will increasingly require quantitative methods for evaluating this influence. Previous studies have already shown the potential of numerical groundwater models to characterize HEF and compare restoration scenarios. Although various sources of uncertainty are acknowledged, the potential effect of small streambed structures (or microforms), such as grains or ripples, embedded in channel-unit scale structures (or macroforms), such as riffle-pool sequences, is commonly ignored. Here, a simple conceptualization through a 2-D vertical model is used to test whether (i) ignoring microforms in groundwater models at the macroform scale can impact estimations of residence times; (ii) microforms can influence HEF patterns driven by macroforms; and conversely (iii) the uncertainty of head measurements in stream piezometers can affect our understanding of HEF patterns. Results show that (i) residence times and flux estimations can be strongly affected by the modeller's choice to represent microform-induced HEF or not; (ii) the interaction of the microform and macroform scales can induce various subsurface flow patterns; and (iii) the perceived significance of microform-induced HEF is highly sensitive to the uncertainty of in-stream measurements of subsurface heads. Little is known about the relative efficiency of these microform and macroform scales, which are effectively influencing exchange at different depths and interacting with each other. Future studies that consider biogeochemical cycling or streambed ecology should be placed in this context. It is also necessary to find ways of including this source of uncertainty in groundwater models of HEF. Copyright © 2012 John Wiley & Sons, Ltd.