Maybe they should scale your ratings based on your rating pattern. I don't know how to do it mathematically but there must be some way to normalize scores if you have enough data. Anybody knows how to do this?
I’d start with assuming everyone’s experience is normal distributed.
Someone who always voted 3 stars probably had the same experience as someone who always voted 5 stars. I’d try out computing the normal distribution of the rater and see where on the distribution this particular vote fell.