Some of the data files were obtained by optically scanning the original tables from the ApJ, etc. This process is notoriously unreliable as far as character recognition is concerned, especially with the 1991 technology available at the time of version 1.0. Therefore there were some single-character errors in the digitised data, which had to be fixed by hand. In some cases I got tired of doing this, so I only guarantee that the particular columns in the data files actually used by the scripts are error-free, although most data files were, in fact, fully corrected. If you try to use the data in some of the other columns for some of the files, you will not necessarily succeed until you have similarly gone through and corrected the remaining character errors. If you feel the need to do this, you can always look up the original reference for the proof-reading process.