I know that in a Markov chain, $$\mathbf{N} = (\mathbf{I} - \mathbf{Q})^{-1}$$ gives a matrix to calculate the expected number of times before absorption that a particular [transient] state is visited, beginning from a particular state. (where $\mathbf{Q}$ is the sub-matrix of the transition matrix corresponding to the transient states).
Now I am interested in the expected number of times that a particular state is visited ONLY from two other specific transient states (for example, how many times on average I go to state 2 from state 1 or state 3?). I don't want to count indirect transitions. Assume state 4 is the absorbing state in this example.
Shall I add together the corresponding two elements in $\mathbf{N}$ (elements in row 1 col 2 and row 3 col 2)?