1. For one, it has an actual User Interface for moving forward and backward through search results, as well as providing information about your current query on demand (such as the search terms used, the number of results, and where you stand in navigating through these results). Making subsequent queries, as well as cleanly removing the toolbar and highlighted search results are also functions of the UI.
2. It does not break the content of the page as a side-effect of its search-and-highlight action, so multiple searches can be performed, one after another, without reloading the page.
3. It does not naively interpret your search terms as Regular Expressions, and it works only on the content of text nodes in the HTML document, so it is safe and perfectly valid to have both Regular Expressions and HTML keywords in your search terms.
4. It deliberately filters out search results that are contained in hidden, invisible elements on the page (these are more common than one might think, especially in Web 2.0 and designed-for-mobile Web pages).
5. It strikes the delicate balance between search speed and search correctness and relevance.