Search is not about keywords. It is about linguistics.
Semantics is not only linguistics, but semantic pair links. For example "Sally threw the ball." Sally is the subject, threw is the predicate, and ball is the object. This is a semantic pair link, sally threw ball.
Queries that are fully semantic yield better results. For example, "Italian restaurant in St Louis". Not only is this a locale search query, it is fully semantic. "Italian" is a modifier of the subject "restaurant". "in" is the predicate and "St. Louis" is the object. You have two semantic pair links, "Italian restaurant" and "restaurant in St Louis".
Let's use the example, "best fruit pies" as a search query.
You will notice that my example query is not looking for a semantic pair link. Instead, ontologies are relied upon exclusively.
An ontology is a series of links in XML or a database. There are many ontologies that can describe many things. Some common examples of ontologies are phrases, synonyms, names, brand names, locales, etc.
"fruit" is a modifier of "pies" and a known phrase "fruit pies" found in a phrase ontology.
"Best" is also a modifier, however does not add value to "fruit" or "pies". "Best" can be found in an ontology of synonyms and likely linked to the phrase "top rated".
When the query engine parses a search query, it assumes that the most valuable terms appear as written from left to right as people tend to write this way. However, from a semantic perspective, we can see this is not always true. Using our example, the order of importance would be "pies", "fruit" as a modifier of "pies", then "best".
When "best" is used or "top rated" is used, in content, it is indexed, however, because so many people create content that used "best", top rated", or "top 10", the value of these terms is low due to over use.
The query engine first matches exact single terms, then synonyms. However SERP filters will reorder the results based upon many criteria such as page rank, inbound link quality and value, title tag, h1 tag, URL, internal linking, actual content, social engagement and activity, citations, etc. These factors can seemingly bring "top rated" on par with "best".
The reason why people use "best" is two fold. One, the belief that it actually helps, and it does somewhat, and two it describes the content. For example, "worst hotels in Atlanta" also describes the content. In this way, "best" does help.
In SEO you have two tasks for such queries. One is to create semantic matches using ontolgies and two is to allow for fully semantic matches. The reason is simple. Focusing one "best fruit pies" exclusively narrows the number of search queries your site or page can be found by. Fully semantic content allows for a greater potential for your content to be found. For example, "Best fruit pies in Austin" adds locale search. "Best fruit pies sent to your door overnight." indicates e-commerce and overnight shipping.
Next, think of how people would speak a search query into their phone or tablet. What spoken searches return too many results? What spoken searches would better find your site or page? Lean toward mobile search with a balance of desktop search which tends to be more verbose. Between spoken and typed searches, there is a difference and it is important to know how people search for both. Even if you are only interest in mobile search, thinking "desktop" will help you with mobile search. Why? Because search is about linguistics and fully semantic content rises to the top every time.
Without knowing your specific goals, I would recommend not following the lemmings off the cliff and focus on content, titles, meta description tags, internal links, inbound links, that are fully semantic and not keywords. Why? Because search is not about keywords. Getting my drift yet?