It is close to impossible for a short test to give a single number representing fuel consumption for all users of all that type of vehicle.
As mentioned, the more the use of computers and turbos (turbos give more tuning options) the easier it is to tune for a result. If the test methodology criteria are published, then the manufacturer knows what to do to do well on the test, real life performance is something else. .
Those of us who are 'rev-heads' know we can always get a better tune from a vehicle than the manufacturer gets off the production line.
One solution is to not publish the criteria. Now where I am, consumer groups do that for ranking supermarket prices, they publish the criteria after not before they do a survey. It works ok, but this is unacceptable for a government.

The strange rule about averaging across all a manufacturer's products is intrusive and benefits the big players, making it harder for a new entrant.

What is behind it,
is it the need for government to grow, for the public service to make more jobs for itself, or the carbon climate scare scam?
I say, all of those, they reinforce each other.