Well, according to the algorithm the +DI & -DI crossing is just the zero crossing of the DX, so you should be able to use those strategies. That's probably the reason why the TA-Lib author did not care to store +DI & -DI separately.