Coulomb breakup is used to infer radiative-capture cross sections at astrophysical energies. We test theoretically the accuracy of this indirect technique in the particular case of 15C, for which both the Coulomb breakup to ^{14}C+n and the radiative capture 14C(n,{\gamma})15C have been measured. We analyse the dependance of Coulomb-breakup calculations on the projectile description in both its initial bound state and its continuum. Our calculations depend not only on the Asymptotic Normalisation Coefficient (ANC) of the 15C ground state, but also on the 14C-n continuum. This questions the method proposed by Summers and Nunes [Phys. Rev. C 78, 011601 (2008), ibid. 78, 069908 (2008)], which assumes that an ANC can be directly extracted from the comparison of calculations to breakup data. Fortunately, the sensitivity to the continuum description can be absorbed in a normalisation constant obtained by a simple {\chi}2 fit of our calculations to the measurements. By restricting this fit to low 14C-n energy in the continuum, we can achieve a better agreement between the radiative-capture cross sections inferred from the Coulomb-breakup method and the exact ones. This result revives the Coulomb-breakup technique to infer neutron radiative-capture capture to loosely-bound states, which would be very useful for r- and s-process modelling in explosive stellar environments.