If the DL corners are not correctly permuted (i.e. DBL is solved by definition but DFR is not in its spot), then the tracing is done assuming that the corner belonging to DFL is numbered the same as the corner currently in DFL (which say is X). Now, based on the table given in the beginners' tracing section and the CP case we get for the solve, we must locate the corner that needs to be swapped with X. Thus, here, instead of having 3 swapping options, we will have only one.
Once the corner that belongs in DFL and the swap partner of X have been located, they must be set up such that the corner belonging to DFL is either an F, F2 or an F' away from being solved, and the swap partner of X is in UBR. One such possibility is shown in the image above. Note, this must be done while preserving the CP case, i.e. using only R, U and slice moves. Once this set up is achieved, execute the F, F2 and F' to solve the DL corners; this will solve CP as well.
It may happen that the DL corners are not correctly permuted, however the CP is solved (i.e. tracing gives the sequence 12). Thus, we do not have any corner to place in the UBR position and solve as mentioned. In this case, set up the corner that belongs in DFL an F2 away from being solved (as shown in the figure above), while preserving CP; then do F U F or F' R' F' to solve the DL corners while maintaining CP. [One can also set up an F or F' away and do F U' F U' F' or F' R F' R F respectively.]
If the DFL corner is twisted in place, then simply pick another orientation to solve since these cases are rare and inefficient. However, for an x2 y2 neutral solver, once every ~2000 solves, it may happen that all DL corner pairs are permuted correctly but twisted in place. In such a scenario, one could do normal tracing to get CP, and place a corner swap pair in UFL and DFR while preserving CP, then do one of the following depending on the direction of twist: either F2 R’ F R’ F R’ F or F2 U F’ U F’ U F
Average movecount: 4 - 5
Average double moves: ~ 1
These situations are relatively less efficient compared to if DL corners are solved, since we have 3 swap options in that case while we only have one in this case. However, most of the cases with DL corners unsolved can be done 2-gen (i.e. with F-U or F-R moves only) which can greatly help TPS and lookahead. The nightmare cases can however take over 8 moves and thus must be avoided at all costs by simply solving in another orientation.
Again, we can use wide moves wherever necessary to simultaneously solve the DL edge. This is a difficult step to get a hang of, and it is strongly recommended to check out example solves.