“…For any closed subset S of G we denote by P (S) the subspace of P (G) consisting of the probability measures whose support is contained in S. For a subset Λ of probability measures on G we denote by G(Λ) the smallest closed subgroup of G containing the supports of all λ ∈ Λ, and by Z(Λ) the centraliser of G(Λ) in G. For λ ∈ P (G), G({λ}) and Z({λ}) will be written as G(λ) and Z(λ) respectively. For λ ∈ P (G) we denote by N(λ) the normaliser of G(λ) in G. We recall that given a (convolution) root ρ of λ, there exists x ∈ N(λ) such that ρ ∈ P (xG(λ)); furthermore, if ρ n = λ then for every such x we have x n ∈ G(λ) (see, for instance, [11], Lemma 2.2).…”