The Cosmic Infra-red Background (CIB) has a similar energy density to that at UV/optical wavelengths, implying that a significant proportion of star formation over the history of the Universe has been obscured by dust. We investigate the dusty star-forming galaxies responsible for the CIB. For this, we use the latest version of the hierarchical galaxy formation model, GALFORM, which is embedded within the Lambda cold dark matter cosmological paradigm. To compute far-IR (FIR) galaxy spectral energy distributions (SEDs), a simple model for the absorption and re-emission of radiation by interstellar dust is used. Recent interferometric observations have highlighted that the coarse angular resolution of single-dish telescopes used for FIR imaging surveys can blend the emission of multiple galaxies into a single source. Simulating single-dish imaging we show that the model can reproduce the difference between the observed interferometric and single-dish derived sub-millimetre number counts. Additionally, we make the prediction that the blended galaxies are typically physically unassociated. The simulated imaging is also used to show that the clustering of single-dish sources is boosted with respect to the underlying galaxy population. We term this `blending bias', and show that it can lead to the dark matter halo masses of FIR-bright galaxies being significantly overestimated. These galaxies are predicted to reside in halos of masses 10^11.5-10^12 h^-1 Msol, and taking the blending bias into account yields a good level of agreement with halo masses inferred from observed clustering. This is also the halo mass range that produces the bulk of the CIB, as it represents the halos most conducive to star formation in the model. We show that the model can predict the observed average FIR SEDs of main sequence galaxies to a remarkable degree of accuracy over the redshift range 0.5 < z < 4. However, a shortcoming of the simple dust model is that it cannot make accurate predictions for mid-IR emission. To address this, we couple GALFORM with the spectrophotometric code GRASIL to compute UV-to-mm SEDs, which we use to make predictions for future James Webb Space Telescope galaxy surveys.