We study decentralised algorithms for spectrum allocation in multi-channel wireless systems. We model channel assignment as a node colouring problem for a spatial point process, a formulation that was proposed by Ni, Srikant and Wu (2011). Considering the metric of the minimum distance between two nodes of the same colour, they obtained scaling results for random colouring and showed that a simply greedy algorithm is within a constant factor of optimal. We refine these results, improving the competitive ratio for the greedy algorithm, and establishing convergence to a limiting random variable for the scaled minimum distance under random colouring. We also extend their results to consider fading.