Why Are Typical Values Negative or Missing in high_sum Anomaly Detection in Elasticsearch?

Hello, everyone.

I have a machine learning job that analyzes a numeric field where values are always greater than or equal to 0. The analysis function I'm using is high_sum, and the field type is float.

The machine learning job also does population analysis and more than 50% of values are 0.

When an anomaly is detected in Kibana, the typical values for some anomalies are shown as "-" or even with values below 0.

Why does this happen?

Thank you in advance.