1. It seems that it is possible to compare happiness of two different people; ie I can say that giving the cake to Mary would give her twice as much happiness as it would give Fred. I think that's all you need to counter your first objection. You'd need something much more formal if you were actually trying to calculate it out rather than use it as a principle, but as far as I know no one does this.

2. This is a big problem. I personally solve it by not using utilitarianism on situations that create or remove people. This is an inelegant hack, but it works.

3. This is why I said I am a descriptive emotivist but a normative utilitarian. The fact that people don't act in accordance with a system doesn't mean the system isn't moral. I'd be pretty dubious of any moral system that said people were actually doing everything right.

4. Yeah, tell me about it. Right now I'm thinking that a perfectly rational person has no essential discounts, but ends up with a very hefty discount because she can't make future plans with high effectiveness. For example, investing all my money now and donating the sum+interest to charity in a thousand years only works if I'm sure both the banking system and human suffering will last a millennium.

"Utilons don't make people happier" is a weird way of putting things. It sounds to me a lot like "meters don't make something longer." If you're adding meters to something, and it's not getting longer, you're using the word "meter" wrong.

I don't know much about academic consequentialism, but I'd be surprised if someone hadn't come up with the idea of the utilon x second, ie adding a time dimension and trying to maximize utilon x seconds. If giving someone a new car only makes them happier for the first few weeks, then that only provides so many utilon x seconds. If getting married makes you happier for the rest of your life, well, that provides more utilon x seconds. If you want to know whether you should invest your effort in getting people more cars or getting them into relationships, you'll want to take that into account.

Probably an intelligent theory of utilon x seconds would end up looking completely different from modern consumer culture. Probably anyone who applied it would also be much much happier than a modern consumer. If people can't calculate what does and doesn't provide them with utilon x seconds, they either need to learn to do so, ask someone who has learned to do so to help manage their life, or resign themselves to being less than maximally happy.

I have a feeling this is very different from the way economists think about utility, but that's not necessarily a bad thing.