The single values for the similarity curve are calculated from the sensitivity-corrected threshold value and the match factor.
Ratio = log( (1000 - Threshold) / (1000 - Match Factor) )
All values of a peak are shown in the similarity curve. You can view it in the Peak Details window. This similarity curve shows a distribution of positive values (pure data points), and negative values (impure data points) across the peak, where the 0 would represent the threshold limit.