Your figures are wrong.
Look at the HQ version of the race. For a start the bends are certainly more than 100m. The standard for IAAF tracks is 115.61m. On the 2nd and last curves (easier to tell than the 1st and 3rd) Coe takes c 15 sec to run from end of back straight to beginning of home straight, and he's operating at about 0.128 sec per metre, which is a conservative figure as that's only 1:42.4 pace.
15.0 / 0.128 = 117m.
Thus the curves would appear to be exactly the same as any other track.
Now the 3rd bend where he runs wide.
He is already wide of Konchellah as they hit the bell. He runs the next 3.5 secs (approx 27m - again a conservative estimate) about 30cm from the inside of the line separating lanes 1 and 2.
The distance of a track is measured 30cm from the curb, which is the shortest distance and fastest line to run in an 800. The radius of the curve from that line 30cm out from curb is 36.8m. Each lane is 1.22m wide, so the distance from the measured curve to the line separating lanes 1 & 2 is 92cm.
Let's say Coe runs that 25m from the bell 40cm inside that line, or 52cm from the shortest line. This represents a radius of 37.32m
Use in 2 Pi R = 234.48m which we need to divide by 2, as this is for both bends run at that distance. That works out at 117.24m
117.24 - 115.61 = 1.63m This would be the extra distance run if he ran 52cm wide of the measured curve for the entire bend. Clearly he only does this for 27/115.61m (0.233) of the curve. 1.63 x 0.233 = 0.38m.
Not a lot.
But, he then moves out wider to about 10cm inside the line separating lines 1 & 2, and doesn't fully move back in front of the rabbit for almost 7sec. That is a lot more than the 30m you suggest.
I actually timed it as 6.75. That is 52.7m out of the 115.61m of the entire bend, which = 0.456.
10cm from the line with lane 2 means he runs a radius of 36.8 + 0.82m. Lets say 0.8m to be conservative. Thus 2x Pi x 37.6 = 236.24/2 = 118.12m
118.12 - 115.61 = 2.51m. This is the distance extra he'd have run if he'd been 12cm on the inside of lane 2 for the entire curve.
He only runs 52.7m or 6.75sec or 45.6 % of the entire bend in this position.
This is 1.144m extra.
Add the 0.38m and the 1.144m together and Coe runs an extra 1.524m extra
At approx 12.8 per 100m, this mans each 1m = 0.128.
0.128 x 1.524 = 0.195sec.
Again there were several 'conservative' measurements taken in that, so I think rounding it to 0.2 is reasonable.
So Coe's run was more like 1:41.5 on a relatively poor track (compared to the current mondo ones) and with no one pushing him on the last lap.