The journal accepts 70% of papers, so how come it has an ISI impact factor of more than 4? The journal accepts all papers that are technically correct, irrespective of significance. That's why we sent our recent paper on malaria and climate change there after dealing with reviewers who didn't like the work or claimed it not to be important at other journals.
An article from last year says that it is due to the $1350 publication fee. Authors who have the money are probably stronger as they have research grants or work at well-funded universities.* Though as the article points out they will waive the fee for anyone who claims they can't afford to pay. Also impact factors are higher in biomedical areas than others. The typical medical article in an ISI journal gets 6 citations after 2 years, which implies an impact factor of 3. I think another reason is that the journal is open access.
The author, Phil Davis, thinks that PLoS ONE could face a problem if they get a flood of low quality articles chasing the impact factor, which would impose costs on the journal as they would have to reject them and not get paid. At the moment the academic editors are volunteers and so the costs of rejected articles are lower than that of accepted ones. But they could move to a submission fee model in that case and only waive fees for developing country authors or not at all.
In 2010 PLoS ONE's IF increased despite publishing more articles.
(HT: Tom Kompas)
* There is an assumption that an author won't pay the fee from their private funds. We paid the fee using my coauthor's funding from the Wellcome Trust who mandate open access publication.