Hi,
That's why I suggested - split the texts into paragraphs - not into individual lines 
But I guess you could also stay with line by line output and use two dynamic text nodes: one where Start is set to "Append directly" (meaning - append to the last line already output)
and another one, where Start is set to "New paragraph". Then in the node conditions restrict the "Append directly" node to output lines where TDFORMAT = '=' or SPACE and the "New paragraph" node - the lines where TDFORMAT NE '=' and SPACE.
I'm not sure where the FORMAT_TEXTLINES comes into play - you are not calling it directly, are you..?
cheers
Jānis