We present improved methods for computing a set of alternative source-to-destination routes in road networks in the form of an alternative graph. The resulting alternative graphs are characterized by minimum path overlap, small stretch factor, as well as low size and complexity. Our approach improves upon a previous one by introducing a new pruning stage preceding any other heuristic method and by introducing a new filtering and fine-tuning of two existing methods. Our accompanying experimental study shows that the entire alternative graph can be computed pretty fast even in continental size networks.