It is slightly strange that they decided to use a subset of all the available spectra (There are currently 104 in the BAA, 392 in the ARAS and 113 in the AAVSO databases, most of which would meet the resolution criterion though many are likely duplicates) I wonder what the selection criterion was.
