Ah, crap, I’m really sorry @JLC this is all my fault. It should be bars of course… :)
So, increase nsamples now. By changing the priors you would like to see a more uniform behavior from the priors. To me it looks strange that some categories are predicted to have very low counts (6,7,8) and some (0 and 9) to have very high counts (but larger intervals too).
Check also if it’s ok to start with a ‘0’ as your data does. I know that sometimes a count must start from ‘1’.