My guess, based on on previous experiments with wavetables, is that the sample select value should be one above what appears to be the middle (so 7 not 6) and the modulation depth should be just under half the length (in your case, 5.99)
I could be wrong though, the same method may not apply to sample slots.