We present previously unpublished photometry of three Large Magellanic Cloud (LMC) microlensing events and show that the new photometry confirms the microlensing interpretation of these events. These events were discovered by the MACHO Project alert system and were also recovered by the analysis of the 5.7 yr MACHO data set. This new photometry provides a substantial increase in the signal-to-noise ratio over the previously published photometry, and in all three cases, the gravitational microlensing interpretation of these events is strengthened. The new data consist of MACHO-Global Microlensing Alert Network (GMAN) follow-up images from the CTIO 0.9 m telescope plus difference imaging photometry of the original MACHO data from the 1.3 m Great Melbourne telescope at Mount Stromlo. We also combine microlensing light-curve fitting with photometry from high-resolution HST images of the source stars to provide further confirmation of these events and to show that the microlensing interpretation of event MACHO LMC-23 is questionable. Finally, we compare our results with the analysis of Belokurov et al., who have attempted to classify candidate microlensing events with a neural network method, and we find that their results are contradicted by the new data and more powerful light-curve fitting analysis for each of the four events considered in this paper. The failure of the Belokurov et al. method is likely to be due to their use of a set of insensitive statistics to feed their neural networks.