The indie games community in particular is an example where lots of 'outsiders' with very limited access to resources are able to build interesting games that speak loudly and provide experiences people have never seen before. Many of those people cannot afford even an extra $100 to submit a game to a service like Greenlight just for a chance to share it with people (the same complaint has often been lodged against competitions like the IGF), and end up having to beg fans to help them handle it.
For someone earning minimum wage, a $100 donation - which seems tiny to me, used to Silicon Valley salaries - is over a day's worth of take-home pay.
A fee this large skews the effort involved in a Greenlight submission such that it is dramatically in favor of people for whom money does not mean much, and has the unfortunate effect of excluding developers who are perfectly willing to invest TIME into their games but do not have extra money to give away just to get into a public voting competition. The games are not worse just because the authors are poor.
Note that services like Apple and Microsoft's marketplaces at least provide some value for your $99: They host your game, they perform basic quality certification, etc. In this case you aren't getting as much for the expense.
EDIT: Some additional thoughts:
Offloading the fees directly onto charity means that the impact of a potential chargeback (most likely to come from a jerk who is looking to put fake entries up for laughs) is more dramatic than it would be otherwise. Now the bill for a chargeback is potentially being sent to the charity (though I'm sure in this case Valve would choose to eat the fees.)
If the goal is merely to filter out joke submissions, the fee does not need to be remotely as high as $100. SomethingAwful uses a much lower cost for forum registrations which does help to filter out noise.
On the other hand, SomethingAwful is proof that a fee is not sufficient to keep trolls, unkind individuals, stupid people and all-around jerks away from your community. Valve can set the fee as high as they want and bad things will still happen.
The problems that currently afflict Greenlight require a serious approach to moderation and a serious approach to content discovery. Right now their approach to both is lackluster at best: The only way to discourage poor behavior on the part of a game creator is the Report button, while poor behavior on the part of commenters generates zero retribution (the most a creator can do is ban them from their game page). Game discovery is nonexistent as there is a single randomly-sorted list of games that reshuffles itself as you page through, making it impossible to simply browse. Even something as simple as 'people who upvoted this game also liked' is not present here.
At present Greenlight is basically a popularity contest, with a group of around ~15 games currently boasting visitor counts in the hundreds of thousands with the rest of the games on the service sitting around 5 to 10 thousand. If Valve's goal was a popularity contest, it makes the idea of paying money to participate all the more ridiculous.