I was curious about the argument you seem to be making here, so I ran it past a researcher I am acquainted with and he said this (with some minor edits for clar

I was curious about the argument you seem to be making here, so I ran it past a researcher I am acquainted with and he said this (with some minor edits for clarity): In the case of IQ, we can identify the model through the exogenous quantity "relatedness." We know exactly how related monozygotic twins are (100%) and the expectation for dizygotic twin relatedness (50%), so we can use the fact that these quantities are known to figure out other model parameters. Since we know that dizygotic twins are exactly half as related as monozygotic ones in the limit, we know the the genetic effect is going to be equal to 2(r_{MZ} - r_{DZ}) and excess resemblance cannot be attributable to genes, since dizygotic twins are not >50% related (on average). Therefore, we can use it to directly figure out the fraction attributable to genetics, then we can see the excess relatedness as due to shared environments, and the residual as the unshared.