This question on Physics SE about apparent brightness during a solar eclipse has a log-scale graph of light intensity versus time. Parts of it are the shape I'd expect: a nearly-smooth curve going from "extremely bright" to "very dim". However, there's a great deal of apparent noise in the graph, with jumps of nearly a full division in the graph (corresponding to about a two-fold or eight-fold change in brightness, depending on if it's a natural or base-10 log). What causes these jumps?
Eric Pauer, who produced the chart, has an interesting web site that covers this eclipse. It is evident that the weather was not great at the time of the eclipse, with intermittent cloud, and it will have been passing clouds that caused the variations in intensity.
You can see the evidence from this page http://www.pauerhome.com/eclipse/eclipse99/1contact/1contact.htm