There is a difference between SIL (or ASIL, for automotive) determination, which is done for each hazard separately, and the general reasonable risk, which is about “the most exposed person” to all existing risks and I think this article is conclusive in illustrating this. Of course, same safety goal (or safety objective) can apply to many hazardous events. Essential is here to keep in mind that when applying the risk reduction measures for each SIL level, those will, ultimately, reduce the risk below the overall reasonable risk. The trickiest part here is how to determine what is tolerable or reasonable.
Which typical value to take as safety target for an application, let’s say highway pilot (as automotive example)? Should we take the minimum number of accidents on highways around the globe? Surely we’ll do our best in applying the measures prescribed by the ISO26262 for the corresponding criticality, add on top measures from "other technologies" and from "external risk reduction facilities", but when should we stop? I think this aspect will become increasingly important the more semi- and autonomous driving will be deployed, because safety engineers will have to take more care on other aspects than strictly “safety-related systems”.