Only a dozen weeks after I intended to, here is the MIREX Meta-Analysis package I described in an ISMIR article last year! The article, "A meta-analysis of the MIREX Structure Segmentation task", uses the extra information about the ground truth that was released in the 2012 edition of the task to understand the state-of-the-art in music structure analysis. Among other findings, we suggest that most algorithms don't accurately reflect how listener's adapt their analysis to the length of the piece, and that researchers may be overlearning on the Beatles dataset.
You can use this package to re-compute the figures in the ISMIR paper. The code is thoroughly commented and it would be a simple matter to adapt it to run on the 2013 results (the paper analyzed the results of the 2012 task). It could also be used by anyone wishing to similarly investigate the results of other MIREX tasks.