I converted to RE85 fuel, changed injectors to larger ones (from 410cc -> 650cc)
Have been mapping the software for a while and have found all the important maps.
If anybody have any idea to solve my problem that would be nice.

Problem is with starting the engine at half-warm state.
Cold start works good 7C - 20C, hot start 70-90C works also good if done under 15min
but half warm start engine 35C - 70C is problematic (only if the engine has stayed shut for over 20min).
Very bad start if the engine has been shut for full work day (outside aroun 20C, engine usually 30C).
Works very good when idling, very good partial load and also strong full load.

All these maps has been altered:

Fuel
------
KRKTE
TVUB
TEMIN/TEMINVA
FKSTT   <- Added more fuel to cold side part +100% and hot part only around 20%
KFHSTT  <- This is strange, it's a 6x3 map with engine and intake temperature (enrichment factor 0.75 -> 0.50 with 0.25 increments).
Really no effect, as it's for really hot start
KFWKSTAB
KFWKSTN
KFWKSTT
FNSSM   <- More fuel added (around 30%)

Timing
-----------
KFZW1, 2, 3, 4 around 6degrees more
KFZWMS
KFZWMN
KFZWSTTM <- around 3degrees more for cold start

Idle/Throttle
-----------------
KFNLLNST  <- Did not really help
KFWDKSTT  <- This is the latest that I have changed and it had some affect, still experimenting

Yes E85 has it's own ways
I have been playing with E85 for some time already, but first time with porsche V8 ME7.1
Had one Audi TT with 1.8T ME7.1 and running on E85, but that was easy just added +25% trim overall and some more timing 

Here's my FKSTT, I have tried to add 50% more fuel to 40-80C part and lower -50% but no big difference.
When cranking wideband lambda goes straight to AFR 20 and then goes quickly to AFR 10-11, so it seems it's too rich.
Therefore I started to lower the enrichment and open throttle more at start.

I will try to keep this topic alive, because as it turns out I have the same exact problem.

With the start-up factor map posted earlier, the car starts (my econtent is at 61% at the moment) nicely when warm, 70C+, when cold too (I mean I had one very nice attempt at 15C). When at around 25-35C, no matter if just standing there for long time, or after couple of hours of engine cool down, it starts like crap, I mean RPMs climb up to 1300, then drop down to very low, then it oscillates for 3 seconds and then stabilises. Sometimes dies after the first dip. Cannot state with confidence what AFRs are, because the sensor is not always warmed up, and regardless, Zt-2 reboots / goes silent for long enough during cranking anyhow. I tried raising the FKSTT map, no visible improvement, if anything, it is worse.

One thing I did notice in the maps -  KFWKSTT has a visible less steep drop in the factor for the 20-40C range, see the attachment.

EDIT: As the car was sitting there with more or less precisely 30C of coolant temperature, I tried the richer map again, failed to start three times in a row. When the AFR came to, it showed ~13.5 in a non running engine, which suggests overfueling to me. So on the next attempt I brought down FKSTT down from 20C upwards and I got a reasonable start (with hesitation, but not a visible rpm dip). But because of the previous attempts just before, this is not yet conclusive.

I have also tried with richer and leaner FKSTT but really no big difference.
It starts very good even at 7C degrees, but when it goes over 25C degreees the start is very poor.
Ethanol content is now 79% with our RE85 from several stations.

I noticed that playing with throttle start angle KFWDKSTT helped the most on warm start but seems that the ECU
really does not open the thorttle right away (looking at log), it takes some time. So there must be some other map that
adjust the throttle angle when shutting down.

This week I have been playing with TVUB to get STFT/LTFT better, as the low load cruise speed trim and idle trim was almost +15-20%
Now I have got it down to around 5%. This made warm start worse, also cold start is not anymore so good as it was before.

No KFWDKSTT on my ECU, at least not in the reference definition. From what I can tell from logs, the whole throttle operation is torque based from the very start in my case (again, just a visual claim). What I know for sure is that for the first 3 seconds after start (500 rpm crossed) the throttle PID has much more aggressive setting, together with a badly tuned fuel this was a problem for me, it was killing the throttle too much on the first RPM spike and unable to recover from that. And I suspect (see note above) that this PID operation has some adaptation parameters. These are, I guess again, unfortunately RAM rather than EEPROM based, so each RAM reset causes a problem. I am in progress of verifying all this and getting to know for sure, just too little time.