Using "Import corrected transcript" for translated subtitles

I have looked at a few other community posts that deal with a similar issue but have not found a great answer. It seems that the timecodes when importing a corrected transcript have no bearing on what Premiere decides the timing of the transcript is.

I am creating a suite of training videos that are being translated into 12 languages, most of which I do not speak. About 45 videos times 13 languages is nearly 600 videos. I would love to have this be a less manual process.

Here is my workflow: (adapted from this reply )

Step 1: individuals do screen recordings of their training while explaining what they are doing in English.

Step 2: I import the video to Premiere (25.6.3 Build 2), cut tops and tails and run the auto transcript for the English transcript. 

Step 3: Slight editing on the auto transcript in the Text window (I have noticed this is a lot better recently) 

Step 4: Export video with no subtitles (used for importing subtitles later) (GFD5 - NoSub.mp4)

Step 5: Create captions with the "Create Captions" button in the Text window

Step 6: Adjust Captions formatting so it looks right

Step 7: Export English transcript to .txt file from the Captions section (3 dots/Export/Export text file) (I have also exported this .txt from the Transcript section, but the Caption section gives me more exact time codes. I haven't noticed any other difference in the files.)

Step 8: Export English version with subtitles (GFD5 - English.mp4)

Step 9: Send English transcript .txt file to be translated (attached) (GFD5 Transcript - English.txt)

Step 10: Recieve translated .txt transcript with English still attached (see attached) (GFD5 Transcript - Portugues.txt)

Step 10.5: duplicate translated .txt and delete English with duplicated timecodes (GFD5 Transcript - Portugues ONLY.txt)

Step 11: Import the "NoSub" video back into the Premiere project

Step 12: Right Click "NoSub" file and "Create New Sequence from Clip"

Step 12.5: rename sequence to new language (GFD5 - Portugese)

Step 13: Open the Text Panel, then double click the "NoSub" file from the timeline so that the Text panel is looking at the file, not the sequence. Taken from this reply 

Step 14: Import new Language transcript using "Import Corrected Transcript" (3 dots/Import/Import Corrected Transcript) (GFD5 Transcript - Portugues ONLY.txt)

! Step 15: Transcript imports with no regard to the timecode from the file

! Step 16: Create captions on the new sequence (GFD5 - Portugese)

! Step 17: Correct caption timing and spacing based on the translated .txt file with English in it

Step 18: Repeat Steps 12-17 500 more times.

I initially tested this with some 30 second videos and chalked up the slight inconsistencies to my lack of precision with those tests. 

EXPECTED RESULT: corrected .txt files import and words in the new language appear according to their timecodes. I can just click "Create Captions" and export the video without further adjusting timing of subtitles.

ACTUAL RESULT: similar to this post, (or this one) text imports in random blocks that have nothing in common with the timecodes from the transcript, outside of the words being in the correct order

WHAT I HAVE TRIED: 

a. imported the original translated transcript (with both English and Portugese) and deleted the English within Premiere. No change. It seemed to work better with the English already deleted.

b. This is dumb, but my timecodes show with colons and this post had semicolons in their timecode. Find and replace colons for semicolons did not help.

Based on this reply, it seems that Premiere ignores the timecode entirely, which is frustrating. Based on other posts, my options are some complicated JSON thing that would take me longer to figure out than just brute forcing it, or this, which seems to only work if the spoken language and transcript match.

Am I missing something? I have not messed with SRT files or anything because I need the files to be easily editable by the people who are translating for me. We are using local translators instead of AI tools because we need to use  precise language in our training.