Question: Where is Maple mistaken?

Why does the equality fail to hold in case of a positive geometric density (with the location parameter v=1,2,... if in general)? What's wrong: symbolical result or numerical evaluation? Thanks.
X := RandomVariable(Geometric(p))+1; M := Mean(X); A := simplify((X-M)/M, size);
Mean(1-A+A^2/(1+A))/M = (Mean(1)-Mean(A)+Mean(A^2/(1+A)))/M; is(%); subs(p = 1/3, `%%`); simplify(%, size); evalf(%)

Please Wait...