@Ferus: Without knowing more about the problem, and the specific representations you used for GA and DE, and how the fitness score is calculated, it is very hard to tell. In general yes it is possible to have very different results using different search techniques in global optimisation of combinatorial problems.

There are dozens of valid techniques though, and details matter.

I'm not sure if this is a good fit for Data Science Stack Exchange. If you provided more details of your problem, the representation you used and your optimisation algorithms, then it might make a good question on some other Stack Exchange. Not sure which . . .

@NeilSlater Yea, me neither... The problem is finding the longest path in a graph and I'm using a vector of ints (vertices) representing each solution (chromosome) and then the evaluation-method goes through the vector and calculates the longest valid subset of vertices that are connected.