Well, not every innovation is an improvement! This is especially true for Prime. Here are the results of Prime 6 with the old muPad engine still available there, compared to your disappointing results with the new FriCas/Axiom symbol kernel.
Of course the "solution" a=any number b=c=0 is wrong, as a,b and c can't be chosen to be zero (same for x2)!
As Luc said - if you feel like doing so and think it may help, you may consider reporting your results to PTC support as a bug in the new symbolic engine.