timeFilter generating different timestamps in alert compared to dashboard

The SQL generated on a dashboard panel:

SELECT
  timestamp AS "time",
  id AS id,
  cast(value as double) AS "value",
  key
FROM 'gvo-logging'
WHERE key = 'System/Memory Usage' AND timestamp BETWEEN '2022-05-24T13:10:39.428Z' AND '2022-05-24T14:10:39.428Z'

executes fine on QuestDB over PSQL

When using the same SQL in an alert rule, the granularity of the timeFilter is much greater, and causes the query to fail:

SELECT
  timestamp AS "time",
  id AS id,
  cast(value as double) AS "value",
  key
FROM 'gvo-logging'
WHERE key = 'System/Memory Usage' AND timestamp BETWEEN '2022-05-24T14:05:58.160103525Z' AND '2022-05-24T14:10:58.160103525Z'

SQL used in alert query:

SELECT
  timestamp AS "time",
  id AS id,
  cast(value as double) AS "value",
  key
FROM 'gvo-logging'
WHERE key = 'System/Memory Usage' AND $__timeFilter(timestamp)

Grafana v8.5.1 (61fe4a1859)

what is the setting on your timepicker? can you share a few screenshots?