(L) U F' R //EO 4/4
F , D R2 D F2 U D' B //4b2 2e 8/12
U2 F2 B2 U^ R2 D^ //HTR 6/18 
R2 F2 R2 D2 B2 L2 //slice 6-1/23
^=w //+0/23
Backup I found early.

Also had:
(D') F' L U (F' R2 F R' F L) //4c3 in 10. 22 opt but needs 2 HT before first qt so didnt find it.
And then million 2c4 from the 3 move EOs. So probably also like 22 in there but no way I can check all of those.

L' (D' B2 R) //EO 4/4
(D) F' L2 F B2 U' //2c4 6/10
B (B' D2 B L2 U2 F) HTR 7/17
(R2 * U2 F2 D2 B2) //wrong slice 5-1/21
*= R2 F2 R2 U2 L2 B2 L2 U2 //solved 8-5/24
But now after DR we have:
(B' D2 B L2 U2 F' R2 U2 L2 [B2 L2 F2 D2] B')
[] is just a blocky fish! So:
[] = U2 F2 R2 B2 //slice -1/23
and widen 1st and last qt for //+0/23

B D R' D //EO+rzp 4/4
F U2 R2 B2 F' R //4a1 4e 6/10
(U2 B L2 U2 L2 [B' D2 B') //HTR 8/18
(R2 U2 D2 L2]) //slice 4/22
[]= F' R2 F' L2 U2 D2 R2 //-1/21

Other:
D (F' D) L U2 L2 R F //4c4 in 8. Also 21 opt but didnt find that
B' D R' D , B' U2 F R2 F' R //4b2 4e in 10. 23 opt.

Cool solution!

(F) R' U2 F' //EO 4/4
L (U R2 U2 F2 D R') //4b2 4e 7/11
U //reduce to 4a1 4e 1/12
(U2 * L2) // 2/14 one edge missing from setup to U but...
 *= U2 R L' B2 R' L' //insert 3 cycle to get the last edges in place 6-5/15
There are variations of this HTR by inverting the last move of DR and/or using the DR variation ending in D2 B2 D L instead.
So noe after DR we have:
U (R L' B2 R' L' U) //HTR 7-2/16
(B2 D2 R2 F2 L2 B2 R2 U2) //solved 8-1/23

The way I found the finish was kind of nice as well since that is hard to see. It doesn't build blocks nor can you get to FR easily. BUT there is this thing I call "pseudo FR" I have been playing around with.
Here we see that when we do (B2 D2) we get a situation where we have all D edges in FR and all U edges NOT in FR. Also it just so happens that corners are also 1 U2 away from FR. This means we have "pseudo FR". Meaning we can solve U/D layers with FR turns only and then do U2 at the end to fix things since all U edges are the ones that are not in FR yet.
So we can do:
(R2 F2 L2 B2 R2) //solve using floppy moves.
(U2) //fix floppy and it even cancels with normal.

Probably would not have seen the finnish normally but these few HTRs in 16 were the only good thing I had so I looked at them longer than I normally would (partly also because they were cool :D)

F D' //EO 2/2
(R') (B' F' L F2 L) //4c4 6/8 Short DR is bad subset as always...
F R2 F R2 F2 D2 F' R2 D2 B //HTR 10/18 Million similar HTRs here. Surely one will have a short finnish.
R2 U2 B2 R2 //slice 4/22
widen all QTs //+0/22

F2 R (D2 L) //EO 4/4
(D F') (D F2 U R2 B') //2c3 6e 7/11
(U' R2 D F2 L2 F2 R2 U) //HTR 8/19
(R2 D2 R2 F2 U2) //solved 5/24

Only sub 12 DR I found...

R' (D2 B L) //EO 4/4
D , B F' U' F D //2c3 4e with 4 variations but 2 of those are 5qt and two are 2c3. 6/10
R2 F' L2 U2 F^ D2 B^ //HTR 7/17
U2 B2 //2e+slice 2/19
B2 U2 R2 F2 R2 U2 B2 L2 //slice 8-5/22
^=w //+0/22

DR is 21 opt but needs 8 mobe HTR and there were TONS of 6-7 move HTRs from both variations on normal and inverse and with NISS so I didnt even have time to look at any 8 move HTRs.

Only good DRs I had. Next best I had was 2c4 in 11. And this was the last EO I checked.

(F' R) D' L' //EO+rzp 4c6e 4/4
B F' U2 F U R2 D //2c3 7/11
L2 F R2 F'^ R2 F^ //HTR 6/17
D2 [L2 U2] B2 D2 //double slice 5/22
[]=R2 D2 L2 R2 //slice 2/24
^=w //+0/24

Bad solve. Messed up some EOs on inverse and wasted a lot of time on that stuff.
DR is 23 opt but quite hard to find in time trouble.

L' F' U2 B //EO 4/4
L (B2 R2 L D' L) //2c4 6/10
(U' R2 D^ R2 B2 D^ R2 D) //HTR 8/18
(B2 U2^ B2) //double slice 3/21
(B2 R L' D2 R' L) //slice 6-4/23
^=w //+0/23
DR opt from all DRs I had time to look at.

Also looked at:
L' B' R2 F , R' , U D' L D L //4b2 4e in 10, 23 opt but those seem really hard
And the extension
L' B' R2 F , R , L2 U' D L U L //4b2 2e in 11, 24 opt findable with 4qt

(B' U' F' R) //EO+rzp 4/4
(D2 B2 F2 D' F2 @ U' F) //4b2 2e 7/11
U'^ L2 F2 B2 * D'^ //HTR 5/16
L2 R2 //2e+slice 2/18
*= B2 D2^ F2 U2 F2 D2^ //slice +3/21
^=w  //+0/21

There is also a 20 to slice if you insert 
@= F2 D2 F2 U2 B2 U2 but that seems to be +2 unfortunately.

But wow! This gives 20,18,21 = 19.67!! tied PB mo3!

F' L B //EO+rzp 3/3
(L B2 L F2 L' D) //4b2 4e 6/9
(F2 [L B2 U2 R']) //HTR in 14 with 4 variations! 5/14
(D2 B2 L2 B2) //slice 4-1/17
[]= Lw B2 U2 Rw R2 //+1/18

Very trivial 18. The only possible pitfall I can see is that there is another DR in 9 from same rzp so you might find that and not look for more and miss this DR because of that.
I actually found that DR first and it gives 21 but luckily I looked for more...
F' L B' //EO+rzp
(D2 B2 R F2 R' D) //4c3 in 9
(R) D2 F2 L' D2 L' //HTR in 15
[U2 L2 B2] R2 //double slice in 19-1=18
[]= D2 R2 F2 U2 D2 //slice in 20 and +1 for 21

(B' U' D R) //EO + rzp 4/4
(U) (B' D2 F' U2 F2 U) //2c3 4e with variations 7/11
([F' U2 D2 R2 B' D2 B]) //HTR 7/18
(L2) //slice :D 1/19
[]=B' R2 L2 D2 B F2 R2 F //+1/20

Also had this variation heavy set of 2c3 4e in 11 DRs:
D (F' U D) //EO
(R) B2 U2 F' R2 B' L //2c3 4e with lots variations

(U' D2 B' R2 D) //EO+rzp 3c2e 5/5
U2 F2 L2 B' L B' //4b2 2e 6/11
D2 B2 [L' F2 L //HTR 5/16
B2 U2 B2 //double slice 3/19
B2 U2 L2] U2 B2 D2 R2 D2 //slice 8-5= 3/22
[]= R' D2 L2 R' F2 R2 //+1/23

optimal from this DR but 4e variation D2 B2 R2 F' R F has a 21.
I looked at that variation also but I didnt get that far since I found these DRs quite late.

(L2 F U2 F) //EO 4/4
(U') L' R' F2 L F2 D' //2c3 4e 7/11
(R) F2 B2 L2 U2 L^ F2 L^ //HTR 8/19
B2 R2^ D2 F2 D2 //slice 5/24
^=w //+0

I had lots of sub12 DRs but all except this were 4 or 5qt...

Also had this:
(.) //E-slice EO on F/B
(U' R) //rzp 4c2e 2/2
U' D2 B U2 B //DR-2s 5/7
(R' L) //DR-1s 0/7 
But of course that is 5qt so I didnt even try to solve it. Cool DR is always bad subset...

R F' D L //EO+rzp 4/4
(F L2 U2 F) //2c4 4/8 with variations!
(R2 U) U' F2 R2 U'^ L2 U'^ //HTR 8/16
R2 D2^ R2 F2 //slice 4/20
^=w //+0

Pretty slow set of DRs to check since you want to switch (L2 U), (R2 U), (' L2 U) and (' R2 U) from all 4 variations so 16 NISSes but it is DR in 8 so it was worth it to spend lots of time on.

L F' U2 B' //EO 4/4
R , U2 F2 D' {R F2 D2 L} //2c3 2e 8/12
(D' R2 B2 D'^ B2 D B2 L2) //almost a double slice but has parity 8-1/19
Luckily we know we can fix party issue by doing either 
{} = R B2 U2 R or {} = L U2 F2 L
the latter gives 2e2e+slice directly so we do that.
And then 2e2e solves nicely with
(L2 B2 L2 B2 L2 ^^ B2) //slice 6-3 2/22
^=w
^^=Uw' U //+2/24

Almost got bailed out from not finding any good DRs (this and 4c4 in 10 were the best ones) but unlucky slices give 24 in the end. Kind of deserved tbh...

Horrible that I had to use 12 move DR but at least it had 4 variations so it was still propably going to give 23-24. It is 23 opt but not trivial since there were variationsit needs 1 move longer LS that I had and it happens to be +0.

L F2 R //EO 3/3
D' B , D2 F U' F' //2c3 4e 6/9
B2 U^ (D^ B2 U2 F2 D') //HTR +2 ext 7/16
L2 B2 L2 D2 B2 U2 //slice 6-1/21
^=w //+0/21

Had several 4c3 in 10 DRs but used most of my time on this one so I tried all 1-2 move reductions to 4b2.