I'll give this a shot...
The early and late ratings are not meant to be compared to each other. Early ratings should be compared to early ratings and late ratings to late ratings. The ratings are designed so that any horse can be compared to any other horse. Horses that have been racing on different surfaces and at very different distances can and often do meet up in the same race. Because of this, the ratings need to be able to differentiate between a horse that runs fast early at 5f on dirt and one that shows speed at 12f on turf. The same goes for late ratings. Finishing fast in a 6f dirt race isn't the same as doing so in a 9f turf race.
I don't keep pars or averages on these ratings...maybe I should. But theoretically, it should work like this. The example below is for older stakes horses. This is about what the average winner would run early and late.
Code:
Fur Sur Ear Late
6 D 145 115
8 D 135 125
9 D 130 130
10 D 125 135
12 D 115 145
6 T 135 125
8 T 125 135
9 T 120 140
10 T 115 145
12 T 105 155
So, in the example Vigors cited, the 14 horse had a much higher late rating than the others because he had been closing well in turf routes and was cutting back to 6f. Obviously this doesn't always result in the horse being competitive. In this case, the pace was really fast and set up nicely for the closer. That isn't rare when the pace gets extreme, slow or fast. The winner often looks good on the early/late ratings.
Another horse in the same race, the 3, was a good example in the other direction. She had her most success running long on the lead in turf races. But cutting back to a sprint, her early speed rating was pretty low and placed her midpack. She couldn't keep up with the sprinters early, raced in midpack and pretty much stayed there all the way around.
It isn't easy to predict the extreme pace scenarios though. TimeformUS didn't indicate the pace was likely to be fast for the Laurel race yesterday. Sometimes we nail it, but of course not always.