The part sat on the bench, the holes lined up, the flanges met, and the ends went together by hand. That proves the part is internally consistent: made to one drawing, straight enough for the joints to absorb weld shrinkage, not cut short. It does not prove it will close on a car already wearing a complete system.
Most fit complaints arrive as a contradiction at this point. The fabricator has a photograph of the part on a jig; the owner has a part that will not go on. Both are true. The disagreement is not about dimensions, it is about what the part must join and in what attitude.
A jig is built to nominal. The car is not. It has been driven, so the mounts have taken a set and the subframe has settled on its bushes; a repair may have moved a mounting point; and the section this part must meet has already adopted that car's shape. The jig describes a geometry that no longer exists underneath.
Then there is load. With the suspension hanging, the underside sits at a different height and angle than at ride height. A hanger, a flex joint, or a pipe close to the floor measures differently in the two states. A part checked with the wheels in the air has been checked in the wrong attitude.
A section on a bench needs one interface to work. In a car it needs both ends to work at once while the rest of the system stays put. Once the upstream and downstream joints are fixed, length, rotation and offset are constrained together. Small errors vanish on a bench because the other side moves to suit; on the car nothing moves.
That is why the symptom is usually one joint that will not start rather than a part that will not go on. The part seats at one end, hangs at an angle, and the far flange stands open on one side. It is being asked to satisfy two datums set independently, by people who never had to satisfy both.
A part that is wrong for the vehicle fails characteristically: it does not reach, or reaches only when the system is forced, and it fails identically every attempt. A part that is right but wrong for this car fails differently. It goes on once the run is loosened, which says the interfaces accept it and the position does not.
The third case is a part made correctly to a drawing describing a different variant. No amount of slotting fixes that. Confirm what the car actually is against the coverage for the platform before blaming the part, because a facelift, a market specification or a body style can change the underside without changing the badge.
Put the car at ride height, on stands under the suspension rather than under the sills, and loosen every joint in the run, hangers included. If the part seats loose and binds when tightened, the dimensions are serviceable and the fault is alignment. If it will not seat loose, measure the gap at each face.
A straight edge across both faces shows whether they are parallel. Grease on one face, assembled and separated, shows whether contact is even or on one edge. Look for a burr holding the faces apart, and at where the hangers sit. Hangers welded on during manufacture are a hard constraint.
Where the gap is small and even, relieving a hole or a face is legitimate, provided the joint keeps full contact and the fastener keeps a face to sit on. Where the gap is large, the part needs re-setting rather than coaxing. A flange pulled flat by bolts loads the weld behind it and fails later at the seam.
If the car is the odd one and the drawing is right, correct the car or record the deviation and build to it. A one-off made around a bent chassis is fine as long as the drawing is amended. Send the deviation back with the drawing and let the custom work start from real geometry.
