http://www.motorists.org/issues/enforce/studies/tti04.pdf
CONGRATULATIONS.
You finally found a webpage that had a truely independent study. Too bad it does not say what you think it says. In fact, it specifically says:
"The problem of red-light-running is widespread and growing; its cost to society is significant. A wide range of potential countermeasures to the red-light-running problem exist. These countermeasures are generally divided into two broad categories: engineering countermeasures and enforcement countermeasures. A study by Retting et al. (4) has shown that countermeasures in both categories are effective in reducing the frequency of red-light-violations.
(Lists several countermeasures, including)
6. If officer enforcement is determined to be unsuccessful or ineffective, then camera enforcement can be considered."
That does not sound like a blanket condemnation of red-light-cameras to me. Like good engineers should do, though, the study recommends other less obtrusive measures be tried first.
As far as the graph, I have seen that kind of thing before. If you read my original post, I already said that there is a small, but temporary increase in safety when the yellow light is lengthened. However, it does not stay. Within 6 months, the people learn the new limits and the number of accidents goes back to where it was before. I looked and did not see anything about this graph being a long term study.