D'(L2 D' B U')
(R B L2 D2 F)  //10+4b2 6E ext from 9+2
F2 R D2 L  //HTR 14
F2 U D R2 U D' L2  // 18 E +3

other：
D(D' B U')
(R F D2 R2 F)   //9+4b2 4E 
only got 22 with 4qs

B' L D2 R  //4c6e 3pairs
(F D' B U2 F R2 F2 D)  //some long shit

but if you take a look at UD axis after B' L D2 R (F D' B)，some magic happens (which of course I didn't notice during my attempt):

B' L D2 R 
(F D' B') U F'  //9+2c3 also findable with rzp2 to 7c8e
(D B2 L2 U' L2 F2 U' R2 B2 R2 D2 L2)   //Fin 19

17 with extension：
B' R' L2 B2 L
(F D' B') D B   //10+2c3
R2 F2 U B2 U B2 U   //Fin 17

(R U D')
B2 R F B2 R B' R'  //10+4b2 4E
F U2* L2 B'   //pHTR  14
*=U2 F2 U2 F2 U2 F2  6-3 HTR 17
L2 U2 L2 U' D' L2 U D'  //Fin 8-4/21

R F'，(L U)  //rzp32
(L' <R2> U2 R' D R' D')  //11+0
B2 R' F2 U2 F2 R' D2 R2  //Direct Fin  19

inspired by Doug:
(L B D' <U> R)  //extended 223+2p 5/5
B2 L' B R' L  //sq on UBR and more pairs  5/10
(U B')  //EO on FB  2/12
(R' L U' L')   //Obl  4/16
(L U' R U2 L' U R' U)  //Fin  8-3/21

B2 L' B R' L U' R U' [L U2 R' U2 R L'] B U' R' U' D B' L'

[]=R B2 R' B2  -3/18

U' D2 L
F D (F U')  //pDR  7+4b2
B' U2 F' U2 F' L2 B'  //pHTR  14
L2 [D2  //16 L2E2E
U' L2 U2 F2 L2 U2] L2 F2 U' //solved 22

[]=U R2 U2 B2 R2 D2 -1/21

L2 U2 R' U  //88 6pairs
(R' L)   //pDR  6+4b2
(F U2 B')  //pHTR 9
(R2 F2 L2 B2 * D2)  //L3E  14

*=B2 D2 R F2 D2 B2 L U2  8-3/19

sadly it's opt insertion 

F R2 F
R B2 R L2 D' L2 D  //10+2c3 2E
(U2 B2 L D2 L' D2 L' F2)  // 18 2E2E
(F2 D2 F2 D2 F2 D2)  //6-3/21 E 

....D' L2 D' F2 D2 F2 D2...=U' F2 D' R2 U' D' B2 U2  +1

with 2 0 2 bad edges count I was expecting something way more better than 22...

edit：I wrote down F B' D as 3c2e but never checked it on inverse（but 17 not that easy for me to found）