In the paper "Subcubic equivalences between graph centrality problems, APSP and diameter", it is shown that the all pairs shortest path problem (APSP) and the problem of computing the betweenness centrality for a vertex (BET) is equivalent under subcubic reduction. That is, APSP has a truly subcubic algorithm, if and only if BET has a truly subcubic algorithm.
In the reduction from BET to APSP, it is assumed that shortest paths are unique. I wonder why this assumption makes sense. Does it mean that if this assumption does not hold, then BET and APSP are not equivalent any more?