Ok another long answer!
First off let me say that I have not created an FEA modeling since leaving university in 1992! But the laws of physics have not changed since then so the loads and reactions ought to be the same, but he way that they are modeled in FEA may be different. There are several people posting on here who do currently model in FEA.
That said what I did way back was introduce material density and a mass for the motor. Michael Costin (Race Car Chassis Design) does something slightly different but then he was doing a hand calc. Basically he turns a dynamic load problem into a static problem by replacing masses with the reactions produced by accelerating them - so no need to introduce any high density objects.
I've sketched a very simple model, to make the concepts clear. Starting simple and getting more complicated: Replace the car chassis with a single beam simply supported at each corner. Then place a mass on the chassis - in this case a purple engine. The engine is bolted to the chassis at each corner.
Now if the whole chassis hits a bump which accelerates both front wheels upwards at 2G then we can calculate the force introduced at the front of the chassis as being 2 times the load born by the front wheels.
If you simplify the effect on the chassis to an upwards acceleration (ignoring the tendency of the chassis to rotate about its CG) then you can find the resulting upwards acceleration of the complete car from F=ma
Knowing the mass of the purple engine you can find the reactions fed into the chassis by the engine at each mounting point. As the whole chassis is assumed to be accelerating upwards the force at by the rear wheels is some value less than 1G times the load born at the rear.
You would then calculate the loads imposed on the chassis beam by the various input forces and reactions. I would think that you could apply this same approach to an FEA model.
If you now replace the beam with a pair of truss frames running down either side of the car from front to back and picking up the engine mounts. Half the vertical loads are taken in each truss frame. So you can calculate the loads in each member - and because his simple model has only considered vertical acceleration all forces are limited to the vertical planes. The FEA model should come to a similar set of loads.
Once you've got the FEA to behave like the hand calc then you can start to make the FEA model more like your actual chassis, so changing its shape and adding the other masses.
The allowable stresses depend on the material - for steel you need to keep the stress under the plastic limit (or you'll have permanent deformation). For aluminium you need to keep the stress beneath the peak stress allowed for the design life that you seek (i.e. 5 million load reversals).
I don't know but suspect that for a steel chassis the desired torsional stiffness will dictate material sections (rather than stress level), but for an aluminium one stresses would be teh limiting factor, rather than stiffness. That is if you use a big enough section to achieve the desired stress levels you'll find that the stiffness is more than adequate.
Let me reiterate; I'm only repeating stuff that I have collated from others. So I'm not saying that if you do it like this it will work. However this is the approach that I'm following and you can see a flaw or a better way please let me know!
Incidentally before some one says you can't make aluminium space frames - yup agreed. If using aluminium you'd design a twin spar (Lotus Elise) or mono because the target stress levels dictate large load bearing sections.