Dijkstra's algorithm is optimal. This means that if there exists several paths from a source node to a target node in a weighted graph the algorithm will return the shortest possible path (or one of the shortest possible paths if there are multiple).
A genetic algorithm is not an optimal algorithm and thus there cannot be a guarantee that it finds the optimal solution regardless of what the authors claim. I would rather assume that their problem they tested on was too easy. They do not provide any charts showing the quality progress so I can't even be sure that they didn't find the optimal solution in the initial population already (which means you can find it easily by chance). The authors claim that "The obtained results affirmed the potential of the proposed algorithm where the convergence was guaranteed to obtain the optimal path in each case." is plain wrong.
Personally, I think the idea to use a genetic algorithm to solve something like a shortest path problem for which there exists Dijsktra's algorithm and also A* is quite ridiculous. I don't think it makes sense to compete with optimal algorithms on small problem instances that A* will efficiently solve anyway. Also the authors' approach of using bit string encoding is not really justified. That's a very complex representation for a list of numbers. The authors state that "The best choice of coding has been shown to be a binary coding [8]" and underpin that claim with a reference from 1975(!), in comparison their article is from 2010. Their test section would be unacceptable for any serious journal: They used a single instance with 10 nodes and 14 edges that they attempted to solve 6 times and then conclude that it is guaranteed to find the optimum.
I would use A* to solve shortest path problems. If the networks are larger you can use bidirectional search which is basically A* from both ends that meet in the middle. This should open less nodes.