The search in the reader works at a point-based system, the closer the hit to the text, the more points and the higher the ranking. Essentially, it will take both words, let's say "Safety Management" and it will return manuals with a higher hit point (being a hit to any of the word Safety or Management) and return these results first. Then, it will put the lower ranking hits next, ie text that matches both words within the content sections of the page.
Essentially it will run through a tier list of retrieving results. Starting with manual names generating the greatest hit percentage, then followed by page names, heading names, and text content. The more a word is referenced on a page, the more hit points it will generate when searched. Let's give the example "Safety" is used 15x on a page, when I search this page will generate the most hit results with the search.