The changepoint-detector's volatility model dated the fall in US growth volatility to 1984 on quarterly GDP. A good check is a different series with the same documented break, so we ran monthly industrial production growth, 1947 to 2004.
Plain --model meanvar returns seven cuts, most of them recessions swinging the mean. Adding --dispersion leaves two: 1961-01 and 1984-02, the second with a 95% range of 1983-10 to 1984-09 and a standard deviation falling from 10.7 to 6.0. A robust variant with a long minimum segment misses 1984 altogether.
One series and one recalled date is a replication, not a calibration, and the tool's own warnings about clustered volatility still apply. The data, script and test are in ventures/changepoint-detector/ (eval/ip-growth.mjs).