I'll agree with the DNF, but other than that, you are taking away all the real possibilities, and that's not really a fair setup for a proper answer to this question.
You are going with the given that:
1) the athlete runs a "relatively smart, even race"
2) that the athlete doesn't bonk
Damn, if only all of us could go into a marathon knowing that those two things were taken care of!
What is "marathon pace" that this athlete has done those runs faster than?
Also, only 10 mile runs at that pace? Hmm. BIG difference between 10 miles and 26.2 miles.
Also, EASY to run a halfway decent half marathon based on 10 mile tempo runs and lots of 20 milers. WAY harder to take that same training up to the marathon level and do as well there.
I once ran a 1:17 half and a sub 33 minute 10k! on the way up to a marathon, and I had runs of 23 miles in there and thought I was doing well and then bonked at mile 21 and ran 3:20.
You haven't given enough information for us to make even a wild ass prediction, and even if you did, you can't take those bad variables out of it.
IF you look at conversion tables, a 1:13 half translates to about a 2:45 (and some say even a little faster) marathon. That is such a wild stab in the dark though...just way too many variables with a marathon.