Richard... I don't know what you mean by saying your plan says 3.11. Do you mean that's your target time? If so, I trust you weren't trying to do the 28K at 3.11 yesterday (unless part of a reputable plan).
If the plan says 3.11, and you're thinking 3:30 would be ambitious... and thinking sub-4 is realistic... you really need to narrow down your target time pretty soon, because that's a massive spread, and you can't train well for all of those possibilities.
As confident as you sound in your shoes, I would look again and again at this possibility. It is just so likely to be the cause in some way. What size are they compared to your normal day-to-day shoes? Have you tried different lacing techniques? Search youtube and there is advice on how to distribute the lacing and distribute the tightness of lacing. Are you using the 'extra' lace holes at the very top of your shoe? These can be used to stop your feet from (imperceptibly) sliding around in your shoe - which could be leading to your toes making slight contact with the front of your shoe - possibly quite slightly, but obviously thousands of times on each 5 mile run.
Failing that, there are syndromes that affect runners that cause parts of their feet/legs to go numb on longer runs. Often these diminish with time/experience. Use the search bar of this forum for the word numb or numbness and you'll find a few threads about it.
Ignoring the Yassoo 800s (which I believe most people ignore for the prediction), I think you draw the wrong conclusion.
Are you suggesting that a 5 minute spread in predictions... from 3 different distances, run on different days, conditions... and presumuably different undulation profiles is somehow an indictment of the Runnersworld predictor?
I think a 5 minute spread suggests a brilliant predictor. In a race (especally a marathon), there are just so many factors that can influence the outcome by that amount... how you hydrate / fuel up, wind, sun, rain, shoelace, toilet stops!, how you sleep / train in the week before.... and of course, how you train in the 14-18 weeks before! All these types of factor influence shorter races too... but tend to be less noticeable.
I suspect that not many people have such a small spread of predictions - but for sure, you just got to run one to really find out.
Based on those predictions, I would GUESS that if you train pretty well, for this your first marathon, that you'll achieve somewhere around the 3:30 mark. Often, people double their HM time and add 20 minutes which tends not to be too far off... but sometimes you need to add more for first timers.
The runnersworld predictor in particular, assumes that you train just as well for a marathon as you do for a 5K... but the reality is that the vast majority of people might have time to train up for a good 5 or 10K but are unable to find time to develop the endurance to be perfectly trained for a half and especially for a full marathon... especially a first one. That's why I think your HM time appears a bit poorer than your 5/10K times and why I'd extend that and say that you'll be quite a lot slower at full marathon than even your HM time predicts.