Remember, this point-to-point idea was basically an after-thought. It's decades away, but the rocket will be launching within the next 3-5 years. So they'll have decades to get it right, and the manufacturing line and much of the infrastructure will be paid for.
But you're right about demand. The biggest economic problem is that it's too big. Each BFR needs 1000 passengers. To be economic, it needs to fly multiple times per day, so each BFR needs to fly like a million people every year. Look around to how many people fly long-haul, and you saturate the existing market VERY quickly.
BUT the world is getting richer. Soon (50-80 years?), there will be 5-10 billion middle class folk in the world, an order of magnitude more than now. And if the time for travel can be cut short like this, then you should have some demand induction taking place. So maybe it'd start making economic sense.
There is a fundamental advantage for BFR versus existing aircraft: SpaceX is able to make Falcon 9s and probably BFRs for about the same cost (a little less, actually) per unit dry mass as a 737 or 777 or A380. But a BFR can do trips that'd last 15 hours in a 777 or A380 in less than 1 hour. That means you can do 10 times as many trips, cutting your amortization time and crew hours by an order of magnitude. But that depends on having enough demand, which is pretty questionable except in the long-term.
Honestly, that's kind of why the BFR idea kind of worries me. It feels like SpaceX analysed the need for orbital space-flights, and realized there was unlikely to ever be the kind demand that would provide the economies of scale needed to do what they want. On the otherhand, there's undeniably a large demand for terrestrial travel, so they came up with a way to use rockets for that purpose. It feels like something to show investors, rather than a workable plan.
But hopefully I'm wrong. Economics aside, its certainly cool, and it would be nice to have an application for terrestrial missiles other than nuclear annihilation.
"with so few flights" is the operative phrase there.
Current launch market is about 60-100/year.
This is why SpaceX invented its constellation: not only does it allow them to make a lot more money than just doing a few dozen launches per year, but it allows them to generate their own demand for launches to justify these crazy Mars rockets. The constellation is for about 12,000 satellites replaced every 4 years or so. Initially 500kg, but Mueller (rocket guy) recently hinted they could get much larger (so let's say 50-100 tons). Therefore, they'll be able to self-generate between 30 and 3000 annual BFR launches on their own. This point to point thing is another add-on market, could be even larger from a launch volume perspective (but, interestingly, smaller from a addressable-market revenue perspective than the constellation). With 3000 flights per year and margin to spare (and the benefit of post-inspection of the rocket in case of a near-miss), you should be about 100-1000x as reliable as today's crewed rockets just to start. Now add point-to-point, and you can do 50,000 BFR flights per year...