Matt Cutts has stated that internal link anchors are much less of a consideration for Google than external link anchors. I would be surprised if you're getting much, if any, value from the text you choose on internal links.
Some months ago MC in a video also stated, that internal anchors are a good way to strengthen the linked pages!
This is one of those discussions I think it's easy to get into the minutia of things and listening to then analyzing every word said about an optimization technique -- I also think sometimes we hit a point of diminishing returns rapidly, but with that said, just for the sake of discussing the minute: I think both of the preceding statements are likely true. Why?
Internal links, much like descriptions, keyword tags, on-page keyword frequency, etc. are very easy to manipulate, which means the larger they are relative to other factors as *positives* in the algo the easier the algo becomes to manipulate. How Can Both Be True?
Internal links could definitely be used to help determine hierarchy of a site, relative importance of page [especially where there are a number of pages related to a given topic]
and even the "depth" of information presented on a given subject.
In those ways internal links could not only strengthen the linked page, but, in-my-opinion, also the linking page. I think internally [meaning intra-site]
the linked text could be taken into consideration to help determine which page from a site should rank for a specific query when there are multiple pages from a site that "fit" a query and *could* rank for it, *but*, I highly doubt "healthy widget juice" in the text of internal links rather than consistently linking "widget juice" is going to push you above the competition for a given query.
Search has moved well beyond keywords and Google has been taking into account the text surrounding a link for years to help determine the topicality of a linked resource.
To do that beyond a few words you really have to move the direction it's obvious they're going now and take into account the *point* of the <p> [or whatever surrounds the link]
rather than the link text itself. Example Site A Links:
<p>Learn how to make [link]healthy widget juice[/link] from scratch.</p>
<p>Here's a [link]self-made widget juice[/link] recipe for you.</p> Example Site B Links:
<p>Learn how to make healthy [link]widget juice[/link] from scratch.</p>
<p>Here's a self-made [link]widget juice[/link] recipe for you.</p>
When you take the surrounding text into account those two sites present exactly the same information. What's linked does not "strengthen" or "change" the content of the linked page, so you really can't treat them any differently in the results based on the text in the links if you want to rank the right page.
To "get it right" you really have to look at other factors like the content actually present on the linked page and if it's more beneficial to visitors on the site linking "widget juice", then that's the page you have to present to them.
They've been running a phrase based system which takes into account the natural frequency of related co-occurring phrases for a given topic for over 5 years and Humming Bird is designed to match "meanings of pages/information in the index" to "intent of a user to find a type of page/information" based on those phrases and related co-occurring phrases and how well they "relate" to a given search query, so I think it might be time for people to quit worrying about the linked keywords and trying to squeeze in the exact right number of repetitions overall on the page, because in the system they're running to find the "right site" and the "right page from the site" the examples above with different linked text have to be considered the same and other factors have to be taken into account. The content of the linked page and likely it's relation to the topic of the page(s) it links to have to be given more importance than what a site owner decides to put in the text of a link.
Keywords are dead; long-live keywords.
Just link what makes sense for visitors and move on.