With such a possibly complex layout, it would help enormously if you could post a sample of your data (with any private data replaced with x's or 9's as necessary) surrounded by the [CODE] /CODE[/noparse] tags. This will help to get your challenge resolved much more quickly.
Additionally, it would likely help to show the tab characters in your sample as the ~ character as you've described, as you won't be able to include tabs in your sample data.
Then we can replace the ~'s with proper tab characters on our side when working toward a solution.
Finally, if you can include a small table of how you'd like the output to appear based on your sample, that should eliminate many further questions on our part.
Anything you want to get working with TABS in the way you describe will need to be created in a consistent manner with the same number of TABS in each row. From your description, as I understand it though I could be wrong, that is not what your extraction is giving you. Indeed it sounds like the report may not be offering that consistently anyway.
If that is the case I would be looking for a template with a trap that extracts all of the lines you want (Only those if possible, otherwise a set of rows that you can filter down to just those lines at some point) and then use Monarch's text manipulation facilities to create and populate the fields you need.
As KeyserSoze says the sample inputs and required outputs would be very helpful for working out the possible approaches.