Hmm, I dont think the indexing model you are doing here is the same as in Indexing using a free random variable
Your interpolation of the data grid model is more similar to a Gaussian process model. Maybe it would be more clear if you can write down your indexing model direct?