Retarding field analysers (RFAs) are one of the only widely accepted diagnostics for measuring ion temperature Ti in the tokamak plasma boundary. Usually a pure, Maxwellian plasma composed of a single fuel ion species of charge Z is assumed, in which case the RFA would measure Ti/Z. Incident impurity ion fluxes are neglected. We study the influence of impurity ions on Ti inferred from a RFA by simulating measurements in a plasma with several ion species. We show that for typical impurity ion fluxes in the tokamak plasma boundary and in the collisionless limit of the kinetic simulations, the assumption of plasma purity causes Ti to be underestimated by as much as ∼20%.