I have a strange problem with text recongition for the file with table-like diagrams. Source .png file, as well as resulting .xml and job data can be downloaded from https://dl.dropboxusercontent.com/u/25356815/ocr.zip
While result file shows pretty good level of recognition, it completely misses "Seafood" section in the right lower part of page. Food names there do not differ much (in terms of image quality).
Is this some kind of test drive limitation, or this is bug?
Hope to hear from You soon,
ADDED: Have more samples of similar type documents with same problems...