Right now NIH requires that people who use NIH money to publish an article be given the right to put it on the internet 12 months later (a very good idea), and the bill removes that requirement. The real issue here is not that people really want to submit to journals that do not allow this (though this is a problem), but that it would allow many journals to change their policies because this would remove a large block that impedes this change.
The basic idea is this: if the publisher puts all of this sweat of labor into making reviewing, editing, and formatting, the journal article, they should get a longer exclusive copyright.
But, most journals do not put in any sweat for peer reviews which are normally done for free, editors usually do their job for zero money plus some cash to cover additional costs to their office of their editing (i.e. extra support staff time). In fact, the journal usually provides near zero value to the process with the exception of publishing the paper copy (which is very useful).
Why this is possible: Usually the market would correct for this (why pay for the expensive journals when there are perfectly good free ones out there?) answer: people really, really want to publish in "top journals" in their field and don't care about all of this other stuff much at all. It's an annoying but present equilibrium.
Was money involved? A recent article at maplight asserted that there was a clear link between the amount of money received by a house member and their sponsorship of HR 801. The first thing to worry about here is this: is it the case that you become friendly to an industry because they give you money, or is the industry acting rationally and giving money to people that they like. This is irreducible unless there is a clear time series or something like taped telephone calls. But the bigger problem with this claim is that the numbers they give fail a simple statistical significance test. Permutation testing is probably the best for this scenario because it is non parametric and yields an exact p-value that only assumes exhangeability between house members under the null (this precludes deals that involve multiple house members, but little else). This test gives a one-tailed p-value of 0.092 which fails at the traditional level of significance of 5 percent.