If you have a backup of your PBL, you're going to be fine. Rename the bad copy of the datawindow so you can use it as reference. Export the datawindow from the backup, see if you can import it into live. If so, compare the working srd with the non-working srd to see what's actually different. 

As far as line numbers go, I'm almost positive that the select statement (line starts with " retrieve=") gets collapsed into a single line and continues from there. Line number will then be off by one (I think line 1 gets ignored). Column number may not match up either since the parser is looking for specific text markers, and if something is missing or improperly spelled, the actual problem will be prior to the shown column number.

Hope this helps.

Line 1 is where it says:  release 17.0;

Count down from there. 

I made another comment but it was mistaken. The forum doesn't allow deleting a post.

thanks all.  I ended up where Brad commented and was able to import an older SRD with corrections

Found an older version of the SRD, ran a compare and added the new bits to the old SRD, then re-imported.  All is good again.  The row/column info from the PB import parser is still a bit of mystery to me.   Counting with or without header and embedded SQL as 1, the reported parse error location still didn't make any sense.