I used to think that (postgraduate) students didn't know how to write good manuscripts because they don't learn how to do so during their undergraduate education. I'm now even more cynical about it -- I believe that we actually actively teach them bad practices instead.
1. Honours students are particularly problematic, since they haven't yet had to write a thesis and are mainly exposed to scientific writing in the form of lab reports. Unfortunately, the way lab reports are written does not resemble any form of document that the students will produce in the career -- whether they go into industry or academia. So what's the point? Sure, they get to do a bit of thinking about the science behind the experiment in the process of writing -- and that's great -- but it does not teach them how to write up science.
2. The biggest problem is that students, regardless of level, are often told to write a certain number of words as part of their academic assignments --'write at least 800 words'. The idea is to make sure that they put in enough work, but the outcome is that you get papers with a lot of filler words and phrases.
I had a masters (by research) student write something along the lines of
"'During the first week(s) of the course, a literature search will be conducted where appropriate sources are gathered'"
instead of
"Targets will be synthesised according to literature procedures"
It would be better to give students assignments where they are told what must be addressed in their essays, and then graded accordingly if they do a bad job of it. If they can complete the task in 400 words instead of 600, so be it.
I now tell my students to read Orwell's "Politics and the English language" in order to learn how to write, as it deals with this directly, and contains some great examples.
3. Postgraduate science writing courses are often geared towards teaching students to write popular science texts, and are often given by people outside their fields. Firstly, they are PhD students and first need to be able to write about science as experts before learning to write pop sci -- they WILL need to do the former, but are unlikely to need to do the latter. Secondly, (bad) pop sci writing often ends up being so devoid of actual information that it's impossible for someone in the field to figure out what it's about -- the way the Nobel prize in chemistry is presented in the newspapers is often so short on detail that it's impossible to know what the discovery is.
4. Role models are hard to find. Few students think that the type of writing they see in social media works for science, but they might not realise that journalists are great sinners along the lines of point 2. Authors of fiction vary in quality, and I find it hard to read modern literature because of how self-indulgent many authors are.
I tell students at the beginning of their postgraduate course to have a look at some of the papers that they read when doing the initial literature search for their research, pick out one good and one bad example, and then try to work out WHY they felt that way.
Lindqvist - a blog about Linux and Science. Mostly.