The PTC code currently pads finalVars, finalMeans, and covariances (and other?) with NaN values to ensure that all amplifiers have the same length for FITS compatibility.  It might be useful to have the read methods unpad these values, so the returned data is whatever length was populated for that amplifier.