It could have been a smarter race

For the first time since running the Des Moines Marathon I've taken a close look at the splits recorded on the race route pads. What they reveal is that I ran a tactically flawed race. Here are the splits:
10k in 56:58, or a 9:10 pace.
Half marathon in 1:59:38, or a 9:07 pace.
19.5 miles in 2:52:32, or an 8:50 pace.
What this means is that for the first 19.5 miles I was on a pace to set a personal record of 3:51:26. Why was this a tactical mistake? I was running injured and my goal was to finish. I didn't think I had more than a 5 percent chance of breaking 4 hours and didn't expect to. Yet, there I was at 20 miles streaking for a PR.
Of course, it didn't happen. My injury, and fatigue, set in over the next few miles and I struggled to a 4:05 finish. I'm happy with that, but now I see that if I'd been smarter, I probably would have been faster.
1. All I really needed to do was run another half just like the first. Start slowly and eventually settle in at a nice 9:07 pace. Had I done that, rather than bust a series of sub-9 miles between 13 and 20, I probably would have stayed stronger and finished much closer to, or under, 4 hours.
2. I should have checked my overall pace. My GPS was set to do that, but I never looked. All I ever did was check my mile splits and the overall time and then do the rough math in my head. That, as I have repeatedly proven to myself, isn't good enough. I don't compute well when I'm fatigued. Had I checked my GPS computations I would have seen that my pace was too fast under the circumstances, and I think I would have decided to slow down. A decision like that, made early enough, might have resulted in a faster overall finish.
Eh, race and learn. I'll do better next time.