Structural model including only formative measurement models

[christophls]

As a first-time PLS user, I am struggling with the following question: Is it possible and meaningful to build a structural model with only formative-measured constructs?

Diamantopoulos/Riefler/Roth (2008) caution against using formative-measured constructs in endogenous positions of structural models. Based on Wiley (2005), they state that there is "no mechanism by which an antecedent variable can influence a formative index". In case that there are only formative indicators, Albers/Hildebrandt (2006) suggest to use multiple regression analyses.

Out of the few structual models I have seen so far, there was not a single structural model with only formative measurement models. Is the above-mentioned argument the reason for this? Are multiple regression analyses the only way out?