I was considering making a change to our practice a while ago because I too was thinking that it was counter intuitive to contour whole heart volumes instead of the left ventricle or LAD. I did some research and found that there are several papers that refer to the topic. I think that there is enough evidence in those papers to conclude that the mean heart dose acts as a good surrogate for predicting CAD and LAD doses and is easier to contour consistently.
The recommended Mean heart dose constraints are proven evidence based doses that predict the risk of late side effects.
The following article is just one example of the work that has gone into this topic http://dx.doi.org/10.1016/j.prro.2012.06.007