Hi - i have a flow that is doing an API call to a service, which responds with a unix timestamp. i'm converting the timestamp to UTC using the following formula:
addseconds('1970-1-1',outputs('Compose_2'),'yyyy-MM-dd hh:mm')
which gives predictable results. I want the results to be in EST and to account for DST so i'm using the convert time zone function to go from UTC to EST, and because i'm filling this data into chart labels and want it to be small, i'm using a custom output format based on the microsoft custom output format notes:

however, sometimes the output isn't close to what it should be:

i'm pulling 48 hours worth of data with timestamps, and, after a few, it reports correctly again - i've taken the unix timestamps and converted them to UTC in excel, then to EST and the final column is the output from the convert time zone action:
Unix UTC EST Output
1655992800 6/23/2022 14:00 6/23/2022 10:00 6/22 22
1655996400 6/23/2022 15:00 6/23/2022 11:00 6/22 23
1656000000 6/23/2022 16:00 6/23/2022 12:00 6/23 00
1656003600 6/23/2022 17:00 6/23/2022 13:00 6/23 01
1656007200 6/23/2022 18:00 6/23/2022 14:00 6/23 02
1656010800 6/23/2022 19:00 6/23/2022 15:00 6/23 03
1656014400 6/23/2022 20:00 6/23/2022 16:00 6/23 04
1656018000 6/23/2022 21:00 6/23/2022 17:00 6/23 05
1656021600 6/23/2022 22:00 6/23/2022 18:00 6/23 06
1656025200 6/23/2022 23:00 6/23/2022 19:00 6/23 07
1656028800 6/24/2022 0:00 6/23/2022 20:00 6/24 08
1656032400 6/24/2022 1:00 6/23/2022 21:00 6/23 21
1656036000 6/24/2022 2:00 6/23/2022 22:00 6/23 22
1656039600 6/24/2022 3:00 6/23/2022 23:00 6/23 23
1656043200 6/24/2022 4:00 6/24/2022 0:00 6/24 00
1656046800 6/24/2022 5:00 6/24/2022 1:00 6/24 01
1656050400 6/24/2022 6:00 6/24/2022 2:00 6/24 02
1656054000 6/24/2022 7:00 6/24/2022 3:00 6/24 03
I am really unsure of where to start to figure out why this happens - thoughts?