There is no technical reason as such, but it would make things a LOT more complex. Games like Unreal Tournament, and in fact, just about every shooting game since Doom have done it that way. The model has no weapon, and the models for the weapons are attached to the hand when the weapon is selected. This is done to allow multiple weapons to be chosen in game.

Now...for RTW, you have a different problem. Quantity.

In order to make a weapon act like it is in the models hand, it needs to be synchronised to the skeletal animation. This means tracking EVERY weapon as a seperate entity, effectively doubling the number of items that have to be tracked during a battle. You also would have to do the same for the shields. Now...we are starting to slow the game down bigtime here.

What CA have done is to make things as simple as possible. By attaching the weapon DIRECTLY to the models skeletal animation, it always moves with the model, and is effectively just a few extra polygons. Swapping weapons around is very simple with the correct tools, and I am sure it really didn't cause CA any hardship to do it that way.

Also...since the models are controlled by pre-programmed skeletal animations, you would still need to make a load of different animation sets for each type of weapon. The model has to move as if it was in the players hand. So...for each WEAPON you would need to design a complete set of animations so they would track the players movements. Most games that allow multiple weapons are just using guns.... one firing animation suits all weapons.

So..in simple terms...it is FAR simpler just to attach the weapon to the model. Less stress on the processor, less problems with animations, and overall simplification of the modelling pipeline. Doesn't help people who can't model to make new units..... but makes it a LOT simpler for those who can