Bug in Atkin Lehner involution for weight 3 forms?

edit

I tried the following example in SAGE

G = DirichletGroup(15)
x=G.gens()
M=ModularForms(x[0]*x[1]^2,3);
S=M.cuspidal_submodule();
f1=S.newforms()[0]
f1.atkin_lehner_eigenvalue()

and got an error saying that the subspace was not invariant! The same happened while trying to compute the L-series, and for level 35 for example. Is this a bug?