The GPL license does indeed restrict some things that could be regarded as freedoms, for example the freedom to distribute a closed source version (distribution only, you are allowed to create a closed source derivative as long you keep it to yourself).
It's a bit like "you have the freedom to do anything you want apart from restrict another person's freedom"
In software, however, they are not arguing for liberty and the rule of law, they're arguing for true freedom.
I guess they have drawn a line in the sand and said that any less free is unacceptable, however, more free is also unacceptable.
I believe Stallman originally wanted to call it "libre software" but there wasn't an equivalent word in english so instead he went with "free software" and thus the confusion was born.
The line in the sand the GPL draws is specifically for the purpose of preventing closed versions of open source software (which would be against the original authors wishes). There are of course more liberal licenses such as the MIT license which do allow you to do this.
The problem with public domain is that outside of the US, in some countries "public domain" is actually more restrictive than e.g. the MIT license. So if you want to help users worldwide, you either need to research local laws or use a simple US-law-based license.
The benefit of the GPL, as I like to phrase it, is that it chains its users with freedom whether they like it or not. It's a very viral license and infects a lot that it touches. (Some businesses are even worried that the GPLv3 is vague enough that "distribution" could mean over the web, which would make the AGPLv3 redundant.) If you look at Stallman's free software song, particularly the part that goes "When we have enough free software/at our call, hackers, at our call/we'll kick out those dirty licenses/evermore, hackers, evermore", you can interpret that as saying once enough software respects user freedoms by default, then yes, we don't need the concept of licenses anymore to force it upon them. The GPL uses copyright as a weapon, as do the Copyleft folks, as do the MIT/etc license folks. As a side quip, this is also similar to basic Communist philosophy: you just need totalitarian control for a little while to set everything up, then the government can dissolve itself and everyone will continue working together.
The MIT/etc. license folks subscribe to the idea that the more things you can do, the more freedom you have. While this is true as a single developer, if you again take into account the end-user freedom then the utilitarian argument can go either way. A program licensed under the GPL makes sure all users of that program and its derivatives are free, regardless of its popularity. A program licensed under the MIT leaves the door open for someone to take that code and put it into another program that's licensed under something restrictive, and all of that program's users are not free. However, some programs don't become popular precisely because of the (A)GPL, whereas an MIT project that's used in a popular project (regardless of that derivative project being MIT) often becomes popular itself.
So for the GPL, the total freedom utility is just num_users for the program and all its derivatives. For MIT, if a non-free user is as bad as a free user is good, it's num_users_under_free_license - num_users_under_nonfree. You can see what needs to be satisfied to choose one over the other. There's also a factor involved for profit if you or someone else wants to make money on the software without the risk of legal source-or-binary redistributions.
I think Stallman recognizes the tradeoffs as well, I've read somewhere he doesn't like but accepts the idea of software (such as video games) initially launching as proprietary while it's in its peak profit window (first few weeks for games, traditionally, though that is changing with DLC), and then later relicensing as open source. Same thing with dual licensing.
As soon as someone stands up and tells me how I can or cannot use something, freedom is gone, no?