Cache-conscious data structures are more efficient than cache-oblivious data structures. While the latter are interesting from a research perspective, in practice the cache sizes are known and can be optimized for.
Would you use a cache-oblivious datastructure that's 10x slower than its regular counterpart? Cache sizes don't change every day, I don't see the point.
In general there are lots of datastructures that are wonderful on paper but whose constant time factors make them infeasible in practice.
irrc there is only a constant factor of time difference between the two.