The 2+2=5 thing doesn't work for this discussion because whether or not it's "true" is entirely dependent upon how we've defined it ahead of time. We've created our own arbitrary system where 2+2=4 by definition, so this is one thing we are absolutely certain of--but only because it's a result of our own theoretical constructs.
If I name my dog Rover, then I'm absolutely certain that his name is Rover, but only because I predefined the system that way. This doesn't really apply to our epistemology discussion here.
Saying that we can't know that 2+2=4 is like saying that we don't know if the word "and" really has three letters. We do know that because we decided what a letter is and what having three of them means.