Ok, I have finally figured it out on my own. In order to generate the optimal mapping for the nodes of two trees, with m and n nodes respectively, you need to do some backtracking in the forest tables.
For each field (x, y) starting with (m, n) of the (m, n) forest distance table, do the following: if the minimum was obtained by summing up the field (x', y') and the edit / delete / insert costs, then write down the mapping and go to the field (x', y') of the current forest distance table. On the other hand, if the minimum was obtained by summing up the field (x', y') from the current forest distance table and the field (tx, ty) from the tree distance table, then go to the field (x', y') from the current forest distance table AND to the field (tx, ty) from the forest table corresponding to the tree (tx, ty). You now need to continue the backtracking in both forest tables individually and collect the mappings from both.