PDFTextStripper calculates the character bounding boxes incorrectly

For a specific file the extracted coordinates provided by a TextPosition stored in a charactersByArticle variable do not match the actual positions of the characters of the content. Some of the rectangles return with zero heights, and others appear shifted on a vertical axis. I am attaching the files illustrating the issue, both the sample file itself and a highlighted bounding rectangles on the 2nd page that mismatch.