Re: "M pace runs"- how accurate are they?
With apologies to wanderingrunner for the combination shameless plug/hijack:
Firstly, you should know that my group precedes this Tues night workout (12 days out from race) with an 8x 1mile (2:00 rest), and a 7x 1mile (2:00 rest) workout one and two weeks prior to the 10x 1mile (2:00 rest), respectively. We run them on a well-calibrated pedestrian/bike path straightaway in a park with a gentle roll and slight hills at either end, as would be typically encountered in a road race. It is VERY helpful to do these with a friend or two who plans to complete ALL the reps.
For warmup, we jog a mile, stretch, then do three or four 50m strideouts. The cooldown is a half-full mile jog + stretching. The workout is always scheduled 2 days after the Sunday long run with a rest day in between. The following days (Wed) are just 5-6 miles easy.
It's a 10x1 mile (2:00 rest) aerobic threshold workout, so ideally you want what your 10 K pace would be on THAT day (with no taper), or round it up to what you think you are capable of producing for ALL the repeats. Your lactate threshold pace is supposed to be what you could only just hold for an hour, so you want to try to be faster than this by as much as possible. For example, my best recent half was 1:24:01 and I was shooting for a sub-3 hour marathon, so I did the 10x 1mile repeats in 6:00 each, plus/minus 2-3 seconds. On a track I would pitch for a little faster (5:50-5:55?).
As for long runs leading up to race day, we do a 20-22 miler three weeks out, followed by 16 and 12 the following weekends, respectively. All of these finish with some marathon pacing, the amount of which depending on your fitness.
Incidentally, I imagine the consensus on this forum would be that your slow marathon times (relative to 10K, half mary) might be due to a deficiency in quality long runs this training cycle (last 20 weeks-ish).