I've seen this sort of problem a couple of years ago.
What you need to do is examine the structure of the text file and see if there is a way of identifying the data you want to extract. You read down through the file and until you come to the bit you want then read the data in as a complete line of text and parse it or, if it's formatted consistently, into a UDT variable set up to hold the structure.
It can be messy. The best way, if you're doing this regularly, is to have the output altered to give you just the data you want. I assume from your question that this isn't an option though.
Hope this helps.
John