In the solow model, if you wanna compare the effect of a higher savings rate, then you compare steady-state output (with no growth) under the new savings rate to a benchmark.
But if you wanna compare the effect of a higher labor efficiency growth rate, then you need to compare the entire new balanced growth path to the old BGP, right?
And I guess this is true for all policy analyses in non-stationary models where the policy affects the BGP.