I do not see where internal links of this type are penalized. However, it highly advisable not to take this tactic for two very good reasons.
1] It would be rightly seen as manipulation and any future, assuming none exist today, algorithm can quickly and easily be applied. This would be a no-brainer especially since the algorithms exist and are applied for inbound and outbound links. The link index is structured as such that it would be an extraordinarily simple change to make a similar penalty happen for internal links without notice. Google does mention this type of activity in their guidelines for inbound (back) link and outbound links.
2] Single term link text has little semantic value. Google is a semantics based search engine and has been since it was a research project in 1997. In the original research paper, it was found that searches against title tags yielded a %2 greater intent match than against content and searches against link text yielded nearly the same result. The research paper goes on to talk about blended searches where the search query was applied to content and the title tag separately, ranked, and then blended to yield a better result. Other papers describe extending the blended search to link text and description meta-tags. All of which, still appears to apply today.
The semantic weighting in order of importance is title tag, link text, then content and content elements. Keep this in mind.
Additionally, content, the title tag, the description meta-tag, and link text are indexed individually. Also keep this in mind.
Please, throw-out the notion that any search engine does direct keyword matches. They do not. Instead, semantic scoring of content, title tags, description meta-tags, header tags, and link text is each placed into a matrix and a matrix of matrices is used to evaluate term weight. Any apparent term match is only incidental to this first step in semantics analysis. Along the way, there is much more semantic analysis that occurs to determine what the page is about which also potentially erases the semantic term weight effect. This goes to matching search intent rather than keywords. When you see a term match in the SERPs, it is not a result of the search process, but rather an effect of being able to easily highlight the terms used within the search query from the index. It should not be considered evidence that Google, for example, matches terms directly or nearly directly. One phenomenon of semantic analysis is that a site can rank high in the SERPs for query terms that do not exist on the page or site nor ever had.
Any link text along with any title tag and header tag should be conversational. It is preferred that a link text be a complete or nearly complete sentence to give semantic meaning to the link. This would include the semantic elements subject, predicate, and object. One example would be, Read our white paper to better understand how atomic dispersion effects combustion. where at minimum, atomic dispersion would be the subject, effects would be the predicate, and combustion would be the object. To further this, read, is the subject (verb), white paper is the object, and how furthers the semantic understanding.
Single term links contain no semantic value whatsoever. In this, the so-called long-tail term matches cannot happen. Keep in mind that long-tail keywords is another fiction exactly in the same manner that keyword matches are a fiction. The point being that there is significantly less opportunity for matching search intent.
These so-called online SEO experts are doing you a serious disservice. Please understand that the keyword chase is parroted by so many SEOs that really have not a clue about how search works. Keyword matching is a myth in modern search engines.