From page 53 of Dragon #318

Text being OCRed
Selecting text - how the OCR engine sees it
Resulting text capture

Unfortunately, this was an attempt to manually OCR from the end of one column to the start of another. The "must shroud his identity and location" bit is from the top of the second column. This means that a search for 'the pirate captain must shroud his identity and location' will fail, since from the OCRed text's perspective, 'captain' and 'must' are not adjacent.

We have identified some - currently unindexed - PDF files where columns aren't recognized at all, these will essentially be impossible to make fully useful. Only one or two word searches would have much chance of success.

From page 21 of Dragon #10

Searching for "Greyhawk"

Text as displayed
Text as captured by the OCR engine - this would normally NOT show up during a search
Search results using experimental regex queries

It's notable, but not yet actionable, that some OCR engines actually capture this text perfectly. In the meantime, it's possible using regular expressions can help fuzz us into a better result.