Thank you very much for the answer, Ekalb!
Unfortunately I can't look at police's motorbike, because I'm not using EFLC dlc but main GTA IV. So I'm using a blank policeb slot, not replacing anything.
I've attached an image with the interface I'm working on. Sirens 3 & 4 are the selected ones, placed on the back of the car. They're 2D objects stolen from the FIB vanilla model and moved to the desired position - same for sirens 5 & 6 (front of the car), which in the FIB model were named 1 & 2.
Now, that's interesting: attaching sirenX_L0 to the chassis_L0 will "block" their default rotary status...
(with "rotary" I mean: in policeb slot, sirens 1 & 2, as isolated objects, rotate on themselves' pivot when in.game activated)
The question is: what do you mean by "attaching"?
Sorry, maybe it is a stupid question, but I'm not used to work with this kind of 3D models (as I usually design with 3D nurbs models with Rhino, meshes are just for exporting & rendering purposes). What I understand is something like a "boolean" between 2 objects. Ora maybe something more simple like just moving layers between the "hierarchy tree"?
That was a stupid question, sorry. I've found the "attach" command and used to link siren 2D model to the chassis_L0. It seems to work just as a boolean operation, great.
Is it necessary to attach also L1 & L2 sub.levels/objects to the chassis? How will the in.game render understand where & what is a light if everything becomes part of the chassis? (I'm asking because I've seen that L0 are high poly objects for near in.game rendering, while L1 & L2 are low poly objects for distant rendering)