Coe ran 1:41.7 and 1:42.3 when he followed decent pace (49.7, 50.8 respectively) and beat the fields by 4 + seconds. The 'best competition' at that time were struggling to break 1:44, so it's not Coe's problem he wasn't able to run against 1:42 men. They didn't exist in 79-81. No one could run close to his pb until Cruz in 84. Despite losing to Cruz in LA, it was actually a lot closer than the final result would suggest. Cruz ran slightly wide for much of the race, about 6m in total on bends, meaning his 1:43.0 could have been more like 1:42.2 in a paced effort where he could have been drafted behind a pacer to 500m or so. Coe ran every bend wider than Cruz, covering at least an extra 10m in the process. That means his 1:43.6 was more like a 1:42.4 in a similar scenario in a paced circuit effort. Moreover, he was also baulked twice in the home straight by Jones when they clashed elbows, hence the delay in Coe getting clear into silver. Without that Coe would definitely have been closer to Cruz at the finish, perhaps 1:43.3. I am not saying that Coe would have won, but in terms of physical ability, there was little between them on the day. Cruz ran a smarter, shorter and better race on the day. But Coe was certainly in 1:42 low shape that day.
Similarly in the 86 Europeans, Coe beat the fastest man in the world that year, Cram, and his winning time of 1:44.5 with a last 200m of 24.8 and last 100m of 12.4, also need to be put into the context of how the race played out tactically. Again, Coe ran every bend wide, even running a large part of the 3rd bend in lane 3. In total he ran even more distance than he did in LA - I would estimate 11m, or 1.3/1.4secs. He ran 3m wide just on the last bend, meaning his last 200m on the rails would have been 24.4 in a 1:43.2 race. This is comparable with the best closing splits of Kipketer in a similar paced race. I recall him running a 1:42.77 in Monaco during his purple patch at the end of the 97 season, in which he had the benefit of running on the rails (thus running exactly 800m) on each bend and being paced and drafted through 400m (52.0 for Kipketer) and 600m (1:18.1 for Kipketer), before launching a tremendous last 200m of 24.6/24.7.
Coe also ran his 3rd 200m in his 1:42.33 WR in 24.8 (according to splits at the time), although video evidence suggests it could have been even faster at 24.6.
The splits given at the time were 24.6, 26.0 (50.6), 24.8 (1:15.4), 26.9.
Very uneven and impressive splits, and I cannot recall any recent (last 50 years) 800m WR where there was such a speed up in the 3rd furlong. Normally the first 200m is the fastest and then each subsequent 200m is a bit slower. Looking at the onscreen clock shown on the video of the race, it shows him finishing in 1:42.1, suggesting that it was running a couple of tenths fast. Thus that would also have been the case at 400m, which showed 50.6 on screen, suggesting that his 1 lap split was actually more like 50.8.
The 200m and 600m splits are not clearly marked on the video footage, but if they were indeed 24.6 and 1:15.4 (as given at the time), then his splits could well have been -
24.6, 26.2 (50.8), 24.6 (1:15.4) 26.9.
Whichever are the more precise splits, it is undeniable that Coe speeded up considerably in that 3rd 200m, not the normal or ideal way of running the most economical 800m.
It is clear that all 3 - Rudisha, Kipketer and Coe (and indeed Cruz) were all capable (based on key career performances) of running a sub 25 secs 200m in a 1:43 flat 800m at their bests.