Why reference articles from and/or submit articles to low impact factor journals?
What would be some reasons why researchers would use papers from a low-impact factor journal, say, a journal with IF = 1?
Researchers use papers that are relevant to their research, regardless of where they've been published. Being published in a less prestigious journal doesn't mean the result is less true - it usually means the paper is seen as less significant, usually because the authors/referees/editors think fewer people will be interested in the result.
But sometimes those people are wrong, and a paper is more significant than it appeared, and sometimes the paper may be obscure but it has exactly the thing you need for the work you're doing.
(Also, IF values vary dramatically by field. The top journals in my subfield have impact factors between 0.6 and 0.7, so an impact factor of 1 doesn't sound low-impact to me at all.)
Wouldn't all researchers concentrate their efforts on starting their literature search from the highest impact factor journals, and then develop their new work on this basis? That seems logical to me and also most beneficial to a researcher's career in academia.
No, they wouldn't. Indeed, if everyone else were doing that, a researcher could get a big advantage by being the only person mining paper published in low impact-factor journals for ideas that other people were overlooking. In my experiences, researchers skim pretty broadly, looking for papers that might suggest new ideas or ways to advance their work, and I've never heard of anyone restricting their reading to the journals with higher impact factors.
The question is unclear, but let me try to answer anyway.
You assume that for any given field of science, there exist journals with high impact factors. This ignores the specifics of individual fields. In very basic sciences, impact factors are usually high, because many researchers from other fields will use the findings in their own for. For example a researcher from the field of neurology will be cited from the people from neurology, psychology, psychiatric medicine, artificial intelligence, computer science.
On the other hand, more niche the field becomes, smaller is the pool of the scientists that could use the results. While a computer scientist may cite the paper from neurology to reference certain brain structure that inspired his algorithm or massive processing cluster architecture, there is basically no way that the neurology scientist would cite the aforementioned computer scientist.
That's why there are huge differences among fields regarding the impact factor of the top journals in the category.
Does this mean that the paper on some novel computing architecture is less valuable? Certainly not. While very few computer engineers working in R&D of new commercial computer systems will be interested in neurology, quite a few of them will read the paper on novel, more efficient architecture of massive computer clustering.
However, since they are end-users, their reads will not transform in citations, while significant proportion people that read advanced neurology paper will publish papers on their own and perhaps eventually cite it somewhere (because they are scientists, in that essence, not end users).
Wouldn't all researchers concentrate their efforts on starting their literature search from the highest impact factor journals, and then develop their new work on this basis?
No. My literature searches start at DuckDuckGo, Google, Google Scholar and maybe my institution's library search. I don't even know the impact factors of the venues of the papers I end up reading.
If the papers (or books, or blog posts, or Stack Exchange Q&A) I find are relevant, I will read them, regardless of impact factor.