This is the approach I too would favour. But does it really matter that much?
The decoherence position is that there is no global collapse, the wavefunction just leaks away to mix with that of the environment in a way that becomes effectively classical in look. So in my terms, this is a standard, locally constructive or bottom-up, view of the causality.
And as a formalism, as a model of reality, this may be all that is needed. It seems a pragmatic way of avoiding the philosophical issues of an actual wave function collapse.
But I too would prefer a more complete story in which the top-down constraints exerted by a decohered environment is also modeled. And I would see the transactional interpretation (with its retrocausality) as being about this expanded view (which sees top-down causality acting from the future even - the lightcone or global spatiotemporal scale).
The question becomes whether the "more realistic" wider view is necessary if the simpler bottom-up approach of "dissipating information with no collapse" does the job. What new predictions would a more complex model, including top-down causality bring here?
I believe it would add more. But I waiting to see exactly what.