Reliability seems to reflect correlated-ness among items measuring a construct. Therefore,we can compute real or population correlation by 1 minus R-squared.Is this a correct interpretation? Does population correlation mean a real correlation measured as 1 - R_squared? Does it mean moderators effects are eliminated?

We can compute real or population correlation(rho) by square-root of 1 minus R-squared.Is this a correct interpretation? Does population correlation mean a real correlation measured as square-root of 1 - R_squared? Here, Reliability seems to reflect extent of correlated-ness among variables/items (measuring a construct).Does it mean moderator effects/interaction-effects are eliminated by the this procedure to reach at true correlation?