The steps for FR+Arrow is as follows:
1) Place the FR edge on the D-layer (0 moves if placed during second square and preserved during BT-Redux)
2) Insert FR while forcing an arrow EO case (6.5 moves)
3) Finish LSE with either a good arrow or bad arrow (9.5 moves for good arrow, 10-13 moves for bad arrow)

Forcing an arrow case is to to reduce the number of algs to learn, and maximizing the number of techniques able to finish LSE,
such as misoriented centers with EOLR, EOFB, possibly DFDB, and maybe even more (EOLRb but just for arrow?)

Example Scramble: U M' U M U R U M2 U' r' U2 M R (Taken from Crafto's example
but uses FR instead of DR as R-layer edge)

Usage of EOFB is helpful for maximizing the number of good arrows (DFDB insertion could also be useful)

Inserts FR while forcing arrow EO case (cancel into EOFB)

Place UFUB on D-layer (bad arrow case)

The movecounts are estimated as follows:

1) For good arrow LSE, placing ULUR on D-layer takes 1-3 moves

2) Finishing 4b takes 0-2 moves + 1 move AUF

3) 4c takes 0-8 moves

Total Moves: 6 + 11 = 17 Moves

4) UFUB requires an extra AUF (but so does 4c for ULUR half the time)

5) Bad arrow requires about 3-4 extra moves to complete 4b

Since this subset is made specifically for the 42 method, the advanced cases starts with a misoriented R-layer
so that applying the inverse of an alg on 8 different arrow cases (F arrow and B arrow with 4 AUFs) results in
8 different EO cases, so for each FR location and orientation, you only need around 9 algs to solve all 32 cases

Original proposal was to always force a good arrow, but obviously recognition becomes virtually impossible
in a speedsolve, hence the average movecount is not significantly optimized

The intermediate cases do the opposite of the advanced cases: by orienting the R-layer beforehand, we reduce
the total number of cases four-fold, as there is now AUF-symmetry, hence fewer algs, but for 42 of course this is inefficient

For the advanced cases, the initial AUF is considered part of the alg, even though you could save ~1.25 moves by only applying the AUF if necessary

To practice L7E scrambles, use this link and put |3,4,6,8,9,10,11||0,1,2,3,5,7,8|||| for the custom scramble

Some of the advanced algs are not optimized movecount-wise due to the great extent of possible cases to exhaust, so likely the movecount could be lowered slightly

For the worst FR-Arrow cases, an alternative L7E method might be more optimal

With EOLR for arrow, we have movecount for 4a + 4b of

For EOLRb for arrow, we have movecount for 4a + 4b of

For EOLRb/EOFBb for arrow, we have movecount for 4a + 4b (AUF included for end of 4c in cases of using EOFBb) of