A lot of this is getting quite off-topic now, but I feel I have to respond to some of these points.
The ONLY way they can be avoided altogether is with flyovers/unders. And other than motorways, they are as rare as hens teeth. A miniscule fraction of a percent of junctions. And that's not changing.
Yes, it is. The current plans place a lot of emphasis on upgrading trunk roads managed by the Highways Agency to dual carriageway and grade separated junctions. If you're interested, there's a list of these projects on the Highways Agency web site.
Widening motorways doesn't create junctions, but it does pour ever more traffic onto the existing roundabouts that most motorway sliproads feed onto. With ever lengthening queues to get on and get off the motorways as a result.
And likewise that is why significant junction remodelling works are going on at some of the major black spots for this, such as the bottom of the M6.
You talk of projects to simplify junctions. And that's true. But equally more and more junctions with traffic lights or roundabouts are created. Every time theres a new housing estate, business park or supermarket built for starters.
But those are rarely heavily congested, nor likely to become so because as you say they are typically there to serve specific local requirements. It's really the main trunk roads that we need to consider if we're comparing the efficiency of road transport with the rail network and potential high speed rail infrastructure.
There is no finite population. The number of cars increases every year.
Perhaps, but any given driver is still driving no more than one of them at once. We're seeing more two-driver households that have two (or more) cars and as a general demographic trend more people are staying single for longer and many of them have their own cars. However, neither of these factors (but particularly the first one) necessarily means all of those extra cars are being driven all the time.
It's also worth noting that with the general hostility toward new/young drivers these days, particularly within the insurance industry, more people are waiting until well into their 20s to take their driving tests, which will reduce the number of (legal) drivers if the trend continues. At least for the next few years, it looks as though this effect is going to more than cancel out the increase in the general adult population that you mentioned. (This is actually one of the stronger arguments for improving public transport provision at the expense of funding improvements to the road network.)
And you are thinking about it in the wrong way completely when you talk of "road space". The only thing space gives predictably you is car parks. The road system is a mostly 2D network. And as such it's limited by it's nodes. The bottlenecks are the junctions.
That is true up to a point, but you are oversimplifying. Traffic engineering can be a surprisingly interesting field, because you get all kinds of perverse-seeming behaviours that actually make complete sense when you consider the actors with their local knowledge making decisions in isolation, but which result in tragedy of the commons kind of outcomes. We see this every time a motorway is congested, when the most efficient way to use the space is to have the traffic slowing down and moving uniformly, but there is always Lane Changing Guy who has to jump around cutting everyone up so he can get there five seconds sooner.
There are also all kinds of circumstances when the modelling these guys use still makes daft assumptions which predictably result in unintended outcomes when implemented. They just spent about half a million pounds "improving" a roundabout on the Cambridge ring road to make it more cycle-friendly, but because they apparently didn't understand the ideas they were borrowing from abroad and didn't implement the whole system, the only reason the roundabout has a lower risk of collisions involving cyclists now is that hardly any cyclists will use it any more. This anecdote is really off-topic now so I'll stop there, but it does illustrate the hazards of using a mathematical model that isn't sophisticated enough when you're predicting how roads will work.