First, I  don't get the individual members with Pivotal package.  But I  had this long response almost finished, but in lieu of that, I simply find the ensembles of little value because they can't nail down a forecast at range and once within 48 hrs, operationals make more sense with their higher resolution.  Blend of models is the better option to ensembles imho because it uses higher resolution solutions in deteriming an average forecast, but it's slow to update and has too much stale data that works to defeat its accuracy. But I  appreciate your post.