Here's something important I've just found out while using a glossary created with an older version of CafeTran.
Older glossaries may still contain the fields "Source term" and "Target term". In projects created with the current version of CT, the codes of the languages / language variants (e.g. "en-GB" for British English) are used for new glossaries and glossary entries. For example, if your working languages are English (Great Britain) and German (Germany), your new glossaries will have the fields "en-GB" and "de-DE" instead of "Source term" and "Target term".
Before adding a term to an older glossary, you MUST change the header of the glossary in order to store them. Choose "Glossary - Edit glossary" and replace "Source term" and "Target term" in the header by the language codes you require. If you don't, your glossary terms will NOT be stored in the glossary and will be LOST. I've just noticed that after adding several terms to an old glossary (and I'm lucky it weren't so many terms so that I won't lose too much time).
Your observation is correct. The latest version of CafeTran supports multilingual glossaries so the program needs a way to detect the correct language column. It does so via the language code in the header. The previous versions supported only bilingual glossaries so the language codes were not needed in the header - the first two columns represented the language pair.
Thanks for pointing it out!