Please double check if you're using the latest version of the Pega OCR component.
Where there are by default turned on parameters responsible for correcting orientation and skew.
Nevertheless please keep in mind that the engine which is used is not able to perform accurate table analysis. In your example there are a lot of tables from which extracted text will not be in an order you may expect.
Mentioned wrong recognition can be related to the quality of processed file or you may not have some fonts installed on your server. Have you installed fonts during engine installation?