I think you have been on the right track with the drafting being the issue. I just did the same ride for the third straight day, each with a different setup, and each with wildly different results.
The ride is 35.3 miles, just a few short climbs totaling 128’ and a fair amount of downhill but nothing that lets me get above 45 mph and even that is very brief. Each time I was joined by 24 bots. My time for all three rides were within 3 seconds of each other so even if i did different parts of the ride not exactly the same, over 35 miles each was almost an identical ride. As a matter of practice, I start out really strong for about the first half mile and then settle in and react as I deem appropriate. Here is a snapshot.
Ride 1: Race, Real Bots, drafting OK. I took the early lead but was caught by a group of 6 bots in just a few miles. Over the next 10 miles two of the bots dropped off the back. Somewhere around the half way mark, the 4 remaining bots just pulled away as normal. No power output could keep up with them and I could not draft off them. I finished at 2.0 W/Kg. The 1st place finisher was at 1.3. Non of the bots finished above 1.6.
Ride 2: Group Ride, Pace Bots, drafting OK. I took off but was quickly run down by a group of 5 bots. They were drafting but I could not draft on them. They slid off and out of site and another group of bots did the same thing. I finished 11th with a 1.9 W/Kg. Of the ten that finished ahead of me, it was an even distribution with 2 averaging the same W/Kg, four had a higher output of 2.0-2.2, and the other four had a lower number 1.6 - 1.8.
Ride 3: Race, Real Bots, NO drafting. I took off as usual and nobody ever caught me. With no drafting, I could see all the bots based solely on their distance behind me. Rather than the usual 5 or 6 all bunched the exact same distance from me, there was almost always a 0.02 mile gap between each. I averaged 2.1 W/Kg, finished first, with the second place finisher averaging 1.4 W/Kg (the best of all the bots) and it finished over a mile behind me.
While I will want to do a lot more testing, it appears that not allowing the bots to draft, since the app is programmed to not let me draft, had a big impact on the race. At least the results of the third ride made sense in the context of the individual performances.