That has also been my experience. If you look at the rudrayamalatantra uttarakāṇḍa in the muktabodha digital library , it was created (with the authors permission) from the file S. Malaviya used to typeset his printed edition. The file was written in a non-unicode font and I had to write a sophisticated program  to convert that to a text file in Harvard-Kyoto (Kyoto-Harvard ?) transliteration. By sophisticated, I mean it was much more than just mapping from one code point to another or combining two or more code points into one or detecting if a vertical line was the bottom part of a ligature or made a preceding letter long etc. etc. . If I remember correctly, I even had to use a computer technique called "recursion" to determine preceding letters.  The "debugging" (making the program correct) was a long process..

Thanks,

Harry Spier