Using this site's declared tyme4ts rules, 15 June 1990 at 22:50 produces 庚午 壬午 辛亥 己亥. The same civil date at 23:10 produces 庚午 壬午 壬子 庚子 because this implementation changes the day pillar at 23:00. The ten-minute difference changes both the Day Master and the hour pillar.
Another implementation may change the day at midnight, convert from a historical time zone, or apply a location-based correction. That does not make one result automatically authoritative. Save the birthplace, local clock time, time-zone and daylight-saving evidence, library version, and day-boundary policy before comparing the two outputs.
Now separate calculation from interpretation. The first record has a Xin-Metal Day Master; the second has a Ren-Water Day Master. A reader may attach very different themes to them, but neither record establishes a trait or event. If the original time is only 'around 23:00', the honest result is two candidate charts and fewer time-dependent claims.
This example is reproducible and falsifiable: another reader can run both inputs and identify exactly where a different result begins. It is more useful evidence than a polished life story whose underlying chart has never been checked.