We present the bolometric light curve of SN 1987A from day 616 to day 1316 after outburst. This light curve is based on infrared J to Q0 broadband photometry obtained at ESO La Silla and a blackbody extrapolation for the far infrared flux beyond 20-mu-m, where no observations are available. Under the assumption that such an extrapolation is correct, roughly 80% of the ultraviolet-optical-infrared flux after day 650 is extrapolated: the reported luminosity is then not only affected by large uncertainties, but it is also mainly based on a theoretical model which cannot be confirmed by observational evidence. A positive detection achieved at 1.3 mm suggests that free-free emission is probably contributing to the energy spectrum. In spite of the uncertainties, our data with the assumption made for the FIR extrapolation clearly show that at least until day 791, the bolometric light curve decreases at the exponential rate reported for earlier epochs, levels off from about day 900 to 1050 above the linear extrapolation of this exponential decrease, and then resumes a slow decrease still above the luminosity predicted by models based on radioactive decays. The implications of such a behavior are discussed, and we show that our results are consistent with an energy input, absorbed and reradiated by dust in the ejecta, from a variable (4 x 10(37) ergs s-1 < L(Bol) < 1.5 x 10(38) ergs s-1) undetected extra energy source, whose origin is still unclear. Our results are also marginally consistent with a constant extra energy input of approximately 10(38) ergs s-1.
