OK, here's a simplified version of the actual situation:
Three points in time: T1, T2, T3.
Three data sets: D1, D2, D3. These data sets represent our knowledge
of the conditions at T1, T2, and T3.
Two models: M1 and M2.
Model M1 starts with initial conditions D1 at T1 and produces output
O2 for time T2. O2 is consistent with D2 to within their error bars.
D2 has smaller error bars than O2, so...
Model M2 starts with initial conditions D2 (instead of the less precise O2)
at T2 and produces output O3 for time T3. O3 is consistent with D3 to
within their error bars.
Does the use of D2 (rather than O2) in model M2 support intervention
between T1 and T3?
No.
(Doesn't preclude it, either.)
Loren Haarsma