Agreed that sizing of effective laminates is only considered to be preliminary... There are many reasons why EL sizing is insufficient to capture all of the nuances of sizing composites. They are really only intended to get you in the right ballpark.
Each EL does have an "accurate" laminate representation of E1, E2 and nu12 in its definition based on a smeared laminate, therefore, it is conceivable to calculate a D12 from the smeared laminate. This will not be the same D12 that you would get from a discrete laminate unless the plies are very evenly distributed.
By the way, this limitation is not just restricted to D12... in fact it is true for all of the D terms. For example, lets say you have a quasi-isotropic laminate, but the 0s are all on the outside and the 90s are in the middle. If you calculate D11 and D22 for this laminate, you would see that D11 > D22. However, if you put all of the 90s on the outside, you will see D22 > D11. If using an EL, you will not see this, you will only see that D11=D22, because the discrete nature has been lost and you are seeing a smeared representation of the laminate.
Phil