I have just looked at the data files from the 2008 and 2013 surveys. I checked v501 and v502 (and v503, which is not very helpful) and looked for other possible variables, including country specific (s*) variables that might help you to separate the possible types of former unions. Like you, I have been unable to find any way to make this separation. In both surveys, only about 3% of women are in the category you would like to separate (if I am interpreting your question correctly). It's my guess that the decision not to make the separation was related to the relatively small number of cases. Sorry about this. It's conceivable that there is a raw data file somewhere that includes the full data from Q601, Q602, Q603, but these are not in the public domain. Unfortunately, DHS standard recode files simply do not include all of the detail in those three questions.