Form identification (which I'm assuming is what you are referring to) is designed to match an entire form, not just parts of the page. If I wanted to identify only part of a form, I would probably use a Zone OCR to identify the document.

 It may be possible to create a 'partial match' master form and use that to identify a form.  That is, make a master form that contains only the selected area you are interested in, and use that to match to the scanned in full forms. I haven't tried this though.