Highy correlated series and seasonal adjustment

[bhaupt]

In doing a little posthumous data cleanup, I was examining two highly correlated monthly series (one is generated as a subset of the other) I hit an odd quandary. The principal series has a strong summer and winter component; the subset series has only a strong winter component. In the decomposition phase of the project, I had seasonally adjusted both, slapped the two of them on a graph together and discussed the implication of the disconnect between the series. However, because the two were adjusted according to themselves, what is failed to be realized is the summer disconnect between the principal and the subset series - and that in essence - we failed to identify and call to attention that summer efforts shown in the subset series were potentially propped up by the principal series.

In essence, when displaying the subset series, it almost feels like I should have applied the seasonal factors from my principal series (subset_psa), and then applied a seasonal adjustment to the result (subset_psa_sa); however, for obvious reasons that also feels wrong.

I guess where this goes is this; when examining highly correlated series, how do you handle disconnected seasonal factors to ensure your client does not walk away with a misconception of a seasonal issue not being present?