The function is working correctly - it's sending the 10 largest fractures to the next event, and in your screenshot the large circular fracture is one of those 10.
To specifically isolate the long fracture meshes, I'd recommend using a Property Test instead, in "bounds ratio" mode...choose a value like 3 or 4, and it'll send out all particles whose longest OOBB axis is 3-4x longer than its second longest OOBB axis (ie, it'll send out long, thin particles, like all of the cracks in your screenshot).
To specifically isolate the long fracture meshes, I'd recommend using a Property Test instead, in "bounds ratio" mode...choose a value like 3 or 4, and it'll send out all particles whose longest OOBB axis is 3-4x longer than its second longest OOBB axis (ie, it'll send out long, thin particles, like all of the cracks in your screenshot).