Firstly, you completely neglect to consider ride sharing, where multiple people are heading in roughly the same direction and so a portion of the trip reduces from 2 legs to 1, ie. the simplest case from your examples would be A->C->B->D(wait)->B->C->A. The more people that use the service, the more likely this is to happen, and the more emissions are reduced. It's the exact same aggregation argument supporting the use of transit.
Secondly, you're only focusing on specific local effects rather than systemic effects. The number of cars on the road would be dramatically lower [1], so we're manufacturing, maintaining, licensing and insuring fewer cars, all of which counts towards reducing emissions.
Then the marginal reduction in convenience means some people would change their behaviour, ie. travel less and/or plan their travel better so they get more done in a single trip.
Frankly, it seems clear that there is no world in which people owning their cars would be superior or even equivalent to a fleet of self-driving cars when it comes to environmental impact, traffic or efficiency.
How? I drive from point A to point B in my car. I'm there for a while, then drive from point B back to point A. Meanwhile, someone else drives from point C to point D, is there for a while, then drives back to point C. Total trip legs: A->B, B->A, C->D, D->C.
Now a self-driving car takes over. It drives me from A to B, goes to C, drives the other person to D, gets me at B, takes me back to A, goes back to D, picks up the other person, and takes them back to C. Total trip legs: A->B, B->C, C->D, D->B, B->A, A->D, D->C.
There is no world in which the second set of trips emits less than the first, given equivalent vehicles. It's simply not possible.
(Note that this is different from the have-everything-delivered scenario. The old way was you drive from A to the store at S, then drive back home to A. Someone else drives from B to S, then back to B. Delivery drives from S to A to B to S, which can be more efficient.)