Specifically, the simulation suggests to me that a flying wing (or any
tailless aircraft) with two single elevons along the whole of the two
trailing edges is not controllable.
This is certainly not true. Many radio-controlled model airplanes and gliders with an all-wing, chevron-shaped configuration and no tail have this exact configuration of the elevons, which are the sole control surfaces.
For example, in the photo above, the elevons are red.
Google "Zagi flying wing" for many more examples.
If the elevons are deflected to pitch up, for example, it generates
negative lift over both wings and the torque (assuming CoG is ahead of
CoL) pitches the aircraft up.
This is not true either. When flying a wing like this, an aft stick input generates a nose-up pitch torque that increases the angle-of-attack of the wing. In the long run, the wing's lift coefficient is increased, not decreased. There is no visible evidence of even a temporary tendency for the aircraft to "settle" (accelerate downwards) when the elevons are first deflected upwards, before the wing has had time to rotate to a higher angle-of-attack, though in theory this tendency must exist to some slight degree due to a temporary decrease in the wing's lift coefficient (but not to the extent that it actually becomes negative). This "settling" tendency might be more significant if the aircraft's pitch rotational inertia were greater.
Conceptually, this is no different than the idea that when we are flying a "conventionally" shaped aircraft, and we move the control stick aft, there must some temporary "settling" tendency (downward acceleration) due to the fact that we've increased the downforce (or in some cases, decreased the upforce) generated by the tail, so the net upward lift generated by the aircraft is now less than weight. Of course, this situation only persists until the aircraft rotates in the nose-up direction enough to increase the angle-of-attack of the wing sufficiently to bring everything back into balance, at a higher angle-of-attack and lower airspeed than were present initially. In actual practice, this "settling" tendency is almost always so negligible as to be undetectable.
It might help you understand what is going on if you were to consider reflexed airfoils. In such an airfoil, the trailing edge of the airfoil is permanently "bent" upwards, contributing a nose-up pitch torque, yet the airfoil still creates positive lift over a wide range of angles-of-attack.
(Source: https://www.diva-portal.org/smash/get/diva2:450812/FULLTEXT01.pdf )
It appears you are getting into trouble by relying on the idea that the Center of Lift is behind the Center of Gravity. In reality, the Center of Pressure is not fixed, and in steady-state flight, the Center of Pressure (of the whole aircraft including the contribution from the elevons) and the Center of Gravity must coincide. When we raise the elevons to create a nose-up pitch torque, this pitch torque can be conceptualized as a forward shift of the Center of Pressure. Then when the wing rotates to a higher angle-of-attack, the Center of Pressure moves back aft until it coincides with the CG again, and the aircraft ends up in steady-state flight at a higher angle-of-attack and higher lift coefficient and lower airspeed than it started with.
A more modern way to think about aerodynamic forces and torques is to view the aircraft as having an Aerodynamic Center that is fixed at the quarter-chord point of the airfoil, plus a pitching moment coefficient that varies according to the shape of the airfoil. With this convention, the lift vector can be viewed as acting at the Aerodynamic Center, but we also have to consider the effect of the pitching moment coefficent. Raising the elevons shifts the pitching moment coefficient in the nose-up direction, as well creating some decrease in the lift coefficient. A symmetrical airfoil has a pitching moment coefficent of zero, and conventionally shaped cambered airfoil with no reflex has a nose-down pitching moment coefficient.