Future interferometers may detect a gravitational-wave (GW) signal from a cosmological first-order phase transition. Reconstructing the underlying particle-physics model from such a signal requires theoretical control over the map from microphysics to the spectrum. For classically scale-invariant extensions of the Standard Model, which generically predict strongly supercooled transitions and strong GW signals, this map depends sensitively on the treatment of quantum and thermal corrections to the nucleation rate. Taking the classically conformal ${\rm U}(1)_X$ model as representative of this class, we scan its parameter space and compare two resummation schemes. The first is a high-temperature effective field theory, matched at two-loop level and including next-to-leading-order corrections to the bounce action, with the nucleation-rate prefactor given by the full one-loop functional determinants. The second is a commonly employed daisy-resummed effective potential, with the prefactor estimated on dimensional grounds. Reconstructing the fundamental model parameters through a Fisher-matrix analysis of injected GW signals at LISA, we find that the daisy-resummation scheme is strongly disfavored, as its theoretical error dominates over the reconstruction uncertainty.