Different story here.

In my case I have an old TM, where the translation was done for an Excel file and at that time the "manual line break" was not set to be a segmentation rule.

So I have segments inside like those:

Now I have pretty the same file, but I did get it via MemoQ from Transit and the same segments are now split by the manual line break, like this:

So what I needed was to produce a translation memory, having the segments shown previously split like the files are now.

In the end of the day I've managed that by exporting that part of TM, which contained the translation for the files from the past to TMX. Then I've located the places where the manual line breaks were (it was a bit tricky, as I did only need to keep those followed by •), then I've done some search and replacing, exported to Wordfast TXT, opened in Excel, replaced all the codes I entered by a manual line break in Excel. Then copied the whole table to Word and back to a new Excel sheet. As the manual line break from Excel is recognized by Word as ^l (manual line break), but this manual line break (^l in Word) is NOT recognized by Excel as manual line break, but as a line break, Excel broke the units as needed to separate lines - and this way, after having copied the Excel to Notepad++ I stored that as tab delimited file and created a new TM out of it in Olifant.

That's the whole story - and it seems to have worked, so might very well be, that I saved me about one day on translation work that way.