Something is missing. It looks like the integration will give you an identity Cl=Cl.
Well, I want to get [tex]C_{l}[/tex], right? That is why I apply [tex]P_{l}[/tex] on the correlation function (1) and integrate - it removes [tex]P_{l}[/tex] and the sum leaving [tex]C_{l}[/tex] * some factor in [tex]l[/tex]. Where do you see a mistake?