How can you be taken seriously if you don't know that those distances at Belmont are around one turn?
The difference should be 5 points on raw figures. If making figures were as simple as using one variant for every race card, life would be easy. Now, I certainly question my share of Beyer figures, but I at least usually understand why the figure given was assigned. Obviously, in the case, the figure maker (Hopkins I assume) split the variant. I haven't looked at the whole card, only the routes, but even if there wasn't a race run after the Peter Pan there is a good basis for splitting the variant.
The reason, quite simply, is Mint Lane. You have a horse that has raced six times, 5 in similarly distanced races. He had run well on three occasions and earned Beyer figures of 86, 88, and 87. All of these were done running loose on the lead, ideal circumstances to record high figures. If you give Casino Drive a 105, you are saying Mint Lane suddenly, while dueling on the lead, rand a 96.
Possible? Sure it is. But what is more likely? With the Beyer of 99 assigned to the winner, Mint Lane gets a 90. I am not saying that is what I will go with, and look forward to going over the whole card soon. The last race will certainly play a factor in my decision.
Like I said, I don't always agree with Beyer or his associates, but it isn't like the figures are done haphazardly and spit out. If you don't like it, you can use BRIS or Equibase figures which are done mechanically. Good luck with that experiment.
Here is a link that contains the Beyer charts for future reference for those interested:
http://www.angelfire.com/la2/wahoo/AB.SPRC.html.