In recent years, dust masses of a few tenths of a solar mass have been found in the expanding ejecta of a number of core-collapse supernovae. How dust forms in such quantities remains poorly understood; theories of dust formation predict lower total masses and much faster formation rates than observations imply. One suggestion to reconcile observations and theory was made by Dwek et al. (2019), who proposed that the dust forms very rapidly, and because of its optical depth, is not initially observationally detectable, only being gradually revealed as the ejecta expand. Observational dust masses at early times would then only be lower limits. Using a large grid of radiative transfer models covering dust masses from 10 −4 to 1 M to calculate both the spectral energy distribution and the emission line profiles from clumpy dust shells, we show that this cannot be the case. Some clump distributions allow dust masses of ∼0.01 M to be concealed in clumps and still predict an SED consistent with the observations. However, these geometries predict emission line profiles which are inconsistent with the observations. Similarly, clump geometries which reproduce the observed emission line profiles with dust masses >0.01 M do not reproduce the SED. However, models with ∼10 −3 M of amorphous carbon can reproduce both the SED and the emission line profiles. We conclude that no large masses of dust can be hidden from view in the ejecta of SN 1987A at early epochs, and that the majority of dust must thus have formed at epochs >1000 days.