Yes, if you limit your area, it's easier. Of course, some areas are harder.

The key is that if you want to shuttle people from their homes to a transfer point near a major road, that's a sucky job for human drivers. Uber and Taxis all have minimum fares which make rides of 3/4 mile unaffordable. The reason is drivers don't want to do this. Due to rush hour patterns the rides are often one way, or there are waits or empty runs between customers. This uses up a lot of human time and is expensive.

Robots cost some money to do empty miles, but much less. And they don't "mind" them, being robots. They will happily sit until a customer needs them, or position themselves in advance of need, and sit there waiting for you, with little extra cost.

Even if the main 15 person van that travels the arterial road is human driven, that can still work out because that's a more regular job, though wasteful if you need to do empty return legs. You don't usually have zero reverse commute, but it can be wasteful as you fill the van in the commute direction, and only half-fill it in the reverse direction -- but every commuter transit system has that issue.

Have not checked the recent numbers -- probably worse with Covid -- but the Dept. of Energy's Transportation data fact book has numbers ever year on efficiency per passenger mile (calculated in part using load factors) for all the modes. About 5 years ago they switched how they did electric transport compared to fossil, so you can no longer directly compare them.

Why is 9 "amazingly" low? Seems right to me. You may suffer from a common sampling bias. More people see a full bus than see an empty one, so one's average impression is that buses have more people than that, because few people see the empty buses, other than the driver and the statisticians.

Of course this varies from city to city as well. 9 is the national average. 23 for light rail cars -- after a major increase in the mid-2010s, but they weigh so much (the light refers to capacity not weight) that they are horribly energy inefficient, as bad as putting everybody in their own Hummer on the worst systems. It's an ecological catastrophe some of the light rail that gets put in. Some of it is a bit better, but none of it can beat an electric car.

A bus is about 20,000lbs and a light rail car can be about 70,000lbs, so the bus with 10 is moving 2,000lbs of bus per person, while the light rail with 23 is moving 3,000lbs of rail car per person.

At the DOT site they show the bus load factor as 10 in 2015, so slightly different calculation than DoE. But getting worse. Can't get how you think the reality is "amazingly" low. I mean, what was your estimate of it before you read the stats?

I have other articles on this but it is a big issue. You have private PuDo which will run by whatever rules the owner wants, though city rules may demand that buildings of a certain size meet certain standards. The good news is that their parking lots, which will get fewer cars, provide available space for this.

Public PuDo will be subject to city policies. Ideally it will be shared as that is most efficient. If I know urban planners, they will try to get exclusive space for their buses and maybe larger shuttles. The best solution would be to share them in the cloud, where the city takes the spaces it manages and hands out reservations to use them, and it can give priority to vehicles with more people.

However, you want to exploit the large supply of small PuDo, namely unused parking spaces, spaces in front of low-use private driveways, spaces in front of hydrants etc. Buses can't use these but there are a lot of them.