Unknown said:

But how to get these X IDs back in the target segments?

The thing is, if you replace the <nbd/> tags already in the source file before it enters Studio, the (now) normal hyphens won't be transformed into these placeholdertags with hte ID numbers by Studio during project creation, Studio will simply see a normal hyphen and display it as such, thus allowing normal term recognition to work.

The only thing is that still remains kind of a problem is that when you want to convert the normal hyphens back into <nbd/> after translation in the cleaned target file, running a search and replace will replace ALL normal hyphens by <nbd/>, which might not be desired in case there were also normal hyphens present in the source file in addition to the <nbd/>.