The number of people that can fit on 1 bus can hit in the ballpark of 50 people, so we're getting a roughly 10x capacity boost here. You lose the "ever-moving system" benefit doing this, but I highly doubt some stopping and starting would lead to an order of magnitude slower speed.
Now, the number of people that can fit on 1 SF BART train is in the realm of 10 cars * 200 = 2000 people per train. The spacing is not going to be on the order of 100x longer than in the individual car case, and the starting-stopping speed is going to be not a bit slower than the bus case.
I'll let you work out the math on that, but it's not even orders of magnitude close. Don't need to be a "visionary" to see that.
The real issue is not highways, but what happens when people try and get off or onto them. It's possible but rare to do this well.
Further, roads and bridges do get used from 2-4 AM when subway systems are generally not operating, so it's the use over the full day that's important not just peak usage.
On page 4, 2006: 80% used a carpool or single occupancy car. 2013: 76% used a carpool or single occupancy car which is counter to their narrative. Further, metro areas are rather large, I commute less than 3 miles on secondary roads in the DC metro area from just outside the beltway to just inside of it and encounter approximately zero traffic.