In unicode you can write characters with diacriticals with either a single glyph or you can combine the character with the diacritical writing it in two glyphs.

This is a problem when one searchs sanskrit etexts.

For example, the letters with diacriticals in the Muktabodha digital library are written with one glyph and as far as I can see GRETIL does the same thing.  But the transcoding utility at  "The Sanskrit Library"  http://sanskritlibrary.org/transcodeText.html

combines letters with their diacriticals in two glyphs.

 So if you used the Sanskrit Library utility to create a transliterated word such as for example: śākti and then searched texts from either GRETIL or Muktabodha for that word your search wouldn't find anything.