Modman is correct, but I want to expand on why. First of all, here's the bow shackle application guide from Crosby which identifies the side loading reductions (page 94; the link starts you at page 92 of a larger document):
Kito Crosby is a world leader in the lifting and securement industry it pioneered, and for which it continues to set the standard.
www.thecrosbygroup.com
When you consider what forces will be experienced by the safety chains during a disconnect, they will range from straight back to straight down to forwards (braking) and every angle in between. As a result, we want the shackle to be able to rotate such that the applied load is straight through the shackle regardless of the angle of the force.
By having the
bow of the shackle on the receiver, with no binding, it can rotate freely as forces dictate. This will always result in a straight load other than at truly extreme angles.
By having the pin of the shackle enter into the receiver, the shackle cannot rotate. The pin and shackle body will bind up on the metal of the receiver as it tries to rotate. As a result, anything other than a load straight back will be a side load of varying degrees.
It's worth noting that in the picture I posted, the shackles are hanging with the pin straight down. That just happens to be the natural resting orientation, but the shackle is not stuck in that position. It will rotate such that the pin is straight back without much force.