But your second recording doesn’t show “the idle animation”, unless you have a very weird idle animation. Look at how the arms and head are twisted. It isn’t a correct aim pose, but it’s also not an unaltered idle pose.
What you’re seeing is classic mis-applied additives. Your additive is assuming a base pose of “arms forward, back bent, holding a rifle” as in Idle_Rifle_Hip. The aim offsets from that are manipulations of the bones relative to that base pose; the arm and head bend up by x degrees, say. So when you feed your idle animation into that, that’s what it does. Takes the idle animation (straight backed, arms at sides, NOT a rifle holding animation) and offsets them by x degrees. It’s adjusting them by the amount that would look right if you were starting with a static aim pose (“Idle_Rifle_Hip”). It doesn’t replace them with the final aim poses, it adjusts them by the difference between those final aim poses and the base pose.
Taking out the layered blend per bone won’t solve the issue in this case, at least not entirely. You need to either replace the AimOffset with a Blendspace (which is non-additive) and use Layered Blend Per Bone to drive that alone, OR be feeding the aim offset with the aim base pose (“Idle_Rifle_Hip”)
