We identify satellites of isolated galaxies in SDSS and examine their angular distribution. Using mock catalogues generated from cosmological N-body simulations, we demonstrate that the selection criteria used to select isolated galaxies and their satellites must be very strict in order to correctly identify systems in which the primary galaxy dominates its environment. The criteria used in many previous studies instead select predominantly group members. We refine a set of selection criteria for which the group contamination is estimated to be less than 7% and present a catalogue of the resulting sample. The angular distribution of satellites about their host is biased towards the major axes for spheroidal galaxies and probably also for red disc galaxies, but is isotropic for blue disc galaxies, i.e. it is the colour of the host that determines the distribution of its satellites rather than its morphology. The similar anisotropy measured in this study as in studies that were dominated by groups implies that group-specific processes are not responsible for the angular distribution. Satellites that are most likely to have been recently accreted show a tendancy to lie along the same axis as the surrounding large scale structure. The orientations of isolated early and intermediate-type galaxies also align with the surrounding large scale structures. We discuss the origin of the anisotropic satellite distribution and consider the implications of our results, critically assessing the respective roles played by the orientation of the visible galaxy within its dark matter halo; anisotropic accretion of satellites from the larger scale environment; and the biased nature of satellites as tracers of the underlying dark matter subhalo population. (Abridged)