I think the idea comes from the fact that fluent adults generally do not need to phonetically decipher words letter-by-letter, but rather recognize them instantly, probably by some combination of shape and context. So memory and context is how good, fast readers actually read usually, thus when you want to teach someone to become a good reader, you may conclude that you need to teach them that. The problem is, of course, that (a) people have not actually deliberately memorized all those words, but rather, these instant brain connections have automatically formed through years and years of reading, and (b) when this strategy fails for an unfamiliar word, a good reader will very much fall back on the letter-by-letter deciphering.
This idea is perhaps more attractive for English educators, than it would be in some other language, because the relationship between spelling and pronunciation is rather more complex (or "loose") than in many other languages.
But yeah, I do wonder if these people just totally forgot how they themselves learned to read, and also, it very much flies in the face of empirical evidence.
I have had some thoughts on school book production, which I may as well post here:
So, here in Germany, there are large, private publishers of school books. These are basically monopolies. They create school books based on the curriculum standards set by the government. There is a "market" in the sense that the teachers at the various schools can choose among several books that are available and approved by the ministry (assuming they get budget approval). Those publishers (e.g. Bertelsmann) are, btw, also behind lobbying efforts to open up more parts of the education system to private interests.
These books get minor updates almost every year, which are insignificant for the most part (lots of changing things around so the page numbers are off). Old editions cannot be bought. This causes confusion among teachers and students. This is, I'm pretty sure, so that the schools will buy a new full set of books every couple of years. The publishers also completely discontinue the books every now and then, and instead publish a totally new books.
The way new books are made is that they hire a bunch of teachers, and tell them to write some chapter of the book, based on the government standards. They try to do this mostly on the cheap, so there's a lot of looking at existing books while barely avoiding plagiarism. The quality doesn't actually improve over time, and there is not really much feedback from the actual teachers using the books.
If, instead, the books were made in some sort of collaborative process, by interested teachers, researchers at universities, and so on, that would improve things a lot I reckon. A bit maybe like Wikipedia or Linux, and ideally with a blanket suspension copyright for educational purposes, so that they can include whatever material they like. Of course the teachers should be paid for this work as well.