One thing I learned a long time ago - it is extremely difficult, if not almost impossible to determine the real cause of a weld defect without putting your eyes on the part, observing the weld process, looking at material chemistry data, looking at thermal history of the part, cutting it up for failure analysis, or some other method that gives reliable data. I can only speculate that since you are using SAW, and it is a deep penetrating process when performed correctly, that you may have some inclusions at the weld fusion line area, or the UT is seeing the interface between the dendritic weld microstructure and the grain structure of the base metal, or the heat input from welding is actually causing microcracking in the heat affected zone, or you have some hot cracking going on in the weld metal. What you really need to do is make some similar welds on a discarded shaft of the same material and have them examined by a local test lab.
Marty