Well, a few things. Firstly, the Falcon Heavy (proposed launch vehicle for Mars One) can put something into orbit for about a tenth of the cost-per-pound required for the space shuttle. Secondly, having usable gravity makes it much, much easier to design and assemble living modules. Thirdly, a potentially significant amount of material can come from Mars itself.
So, yes, we'd certainly still run into expenses, I just don't think you can use the ISS as a baseline and say "...and it only gets more expensive from there." Some things do, but a lot of things don't.
Sure, of course. Scaling up, you have 16 billion for Mir, and 150 billion for the ISS. Roughly ten times cheaper for equivalent mass, although I'm not sure if that includes all of the costs. Even it doesn't, the number is low enough to at LEAST imply that the ISS is far from the cheapest example of non-earth infrastructure.
I'm not an expert, but I'm not sure the difference is as great as you think.
This is undoubtedly the most difficult and expensive part.
Water gets recycled, energy is probably nuclear + solar. No problems there. Food is the big one, but probably not ridiculously so. By my calculations, a lifetime's worth of dehydrated food for four people comes to about 100,000 pounds, which is still a fraction of the total mass we're probably talking about. It can also be sent cheaply as it doesn't need life support and isn't fragile. It can, within reason, more or less sit on top of a dumb rocket and "crash" into Mars. You also don't have to send it all at once, and launch costs will assuredly continue to come down over the years. And that's all assuming that the colonists can't grow any of their own food whatsoever, which is probably not the case.
Yes and no. Even with the space stations and such, we can't get anyone off there in a hurry. If you have a hull breach or a heart attack or anything else that will kill you in less than a day or two, you're screwed, despite being in close communication with Earth. Because of that, everything is already designed with excessive redundancy in mind. I don't see that you'd do anything different for a Mars mission.
