I have to agree with Moraghan in the overall discussion, i.e. in the argument that the composition of the OP's training programme is being almost completely disregarded.
Comparing to my own training / race time predictors leading up to my first marathon (sub-3 attempt), my HM leading up to the race was 1:23, 10 mile time 1:02 (marathon predictors 2:55; 2:53 respectively). I'd not run a recent 10k but post-marathon performances suggested I would have been in about 37 mins shape at best. But my training consisted of 5 runs/wk, max 54mpw and the fastest running I ever did was a weekly tempo run, plus an steadily increasing proportion of MP (or thereabouts) leading up to the race, including a 20 mile race w/ 10M @6:50, which felt very comfortable. Even then, less than ideal conditions on the day saw me fade a bit at the end, with a 3:04.
I'm fairly confident there's no way I could have matched the OP's 36:03 10k prior to my marathon - because I wasn't suitably trained for it. Here's the danger with race time predictors - McMillan equates 36:03 with a 2:49 marathon. So plenty to play with there, it must be on surely? But it also predicts a 1:20 HM... which hasn't happened. So if you're already falling short on the half, who knows what this will tell you about the last 6 miles of a marathon, trying to maintain a pace of <7m/m on tired legs?
After all that, I'm gonna sit on the fence on this one. I don't think we've got enough information on just how much suitable endurance training the OP has done and whether it's enough to minimise the potential fade towards the end of the race. I'll just wish you good luck. 