I can tell you why its rated second.

Neurals are only as good as the data that is used to train them. In the data they used to train the neural black box a few horses with 0 prize money have probably won. This makes the neural network think that 0 prize money is a good thing. Logic on the other hand tells you this is a bad thing.

I've analysed the neurals from what I believe they actually measure and quite a few of them are not very good. You need to know which ones to use. $$$ has always been a bad rating to use. The others to avoid are DLR, JT, TA and TIM. When I say avoid it depends on which races you use them in.