Metrics in HPA not matching Prometheus/KEDA

[mwm5945]

Hi all,
I've created a SO that contains the following triggers:

   triggers:
   - metadata:
       metricName: istio_requests_total
       query: |
         sum(
             rate(
                 istio_requests_total{
                   [tags]
                 }[5m]
             )
         ) /
         count(
           count(
             container_memory_usage_bytes{
               [tags]
             }
           ) by (pod)
         )
       serverAddress: http://prometheus.istio-system.svc.cluster.local:9090
       threshold: "300"
     type: prometheus
   - metadata:
       type: Utilization
       value: "75"
     type: cpu

The above query is basically RPS / Pod. With about 120 RPS, and 2 pods, and If i run the Prometheus query directly on that Prometheus, i get the expected value of about 62. Additionally, if enable the metrics scraping to prometheus, keda_metrics_adapter_scaler_metrics_value yields the same answer, roughly 60. However, the HPA is getting about half that:

Metrics:                                                                                                           ( current / target )
  "prometheus-http---prometheus-istio-system-svc-cluster-local-9090-istio_requests_total" (target average value):  31500m / 300
  resource cpu on pods  (as a percentage of request):                                                              22% (134m) / 75%

Querying the External Metrics API Directly also yields the correct value:

❯ kubectl get --raw "/apis/external.metrics.k8s.io/v1beta1/namespaces/<namespace>/prometheus-http---prometheus-istio-system-svc-cluster-local-9090-istio_requests_total" | jq .

{
  "kind": "ExternalMetricValueList",
  "apiVersion": "external.metrics.k8s.io/v1beta1",
  "metadata": {
    "selfLink": "/apis/external.metrics.k8s.io/v1beta1/namespaces/bonsai-maas/prometheus-http---prometheus-istio-system-svc-cluster-local-9090-istio_requests_total"
  },
  "items": [
    {
      "metricName": "prometheus-http---prometheus-istio-system-svc-cluster-local-9090-istio_requests_total",
      "metricLabels": null,
      "timestamp": "2021-02-05T16:07:24Z",
      "value": "61"
    }
  ]
}

Any idea as to why the metrics is halving itself?

Thanks!