Is there cut off for acceptable Z scores?

I was wondering if there is a limit to what would be an acceptable Z-score. I undertsand a lower Z value the better, but is there a cutoff, say higher than -1, when the clsuters might be consdidered as not being in good agreement withe experimental data?

Use rather the HADDOCK score, the z-score is only a statistical measure of how far is a cluster from the average HADDOCK score of all clusters.

E.g. by definition if you get two clusters, one will have -1 and the other +1 as z-scores.