You'd think context-less OCR would prefer interpreting it as a simple hyphen, since that's the most common dash. Seems unlikely any bias would go the other way.