The pre-explosion observations of the Type II-P supernovae 2006my, 2006ov and 2004et are re-analysed. In the cases of supernovae 2006my and 2006ov we argue that the published candidate progenitors are not coincident with their respective supernova sites in pre-explosion Hubble Space Telescope observations. We therefore derive upper luminosity and mass limits for the unseen progenitors of both these supernovae, assuming they are red supergiants: 2006my (log L/L-circle dot = 4.51; m < 13 M-circle dot) and 2006ov (log L/L-circle dot = 4.29; m < 10 M-circle dot). In the case of supernova 2004et we show that the yellow supergiant progenitor candidate, originally identified in Canada-France-Hawaii Telescope images, is still visible similar to 3 yr post-explosion in observations from the William Herschel Telescope. High-resolution Hubble Space Telescope and Gemini (North) adaptive optics late-time imagery reveal that this source is not a single yellow supergiant star, but rather is resolved into at least three distinct sources. We report the discovery of the unresolved progenitor as an excess of flux in pre-explosion Isaac Newton Telescope i'-band imaging. Accounting for the late-time contribution of the supernova using published optical spectra, we calculate the progenitor photometry as the difference between the pre- and post-explosion, ground-based observations. We find the progenitor was most likely a late K to late M-type supergiant of 8(-1)(+5) M-circle dot. In all cases we conclude that future, high-resolution observations of the supernova sites will be required to confirm these results.