How can this happen and is there a correction?
From what I understood, you mean something like this:
1
00:02:47,000 --> 00:02:47,005
Mr Lee, ik ben Braithwaite.
2
00:02:53,000 --> 00:02:53,005
Ik heb een belangrijke zaak te bespreken.
3
00:02:57,000 --> 00:02:57,005
Thee?
- Erg graag.
4
00:03:03,000 --> 00:03:03,005
Heerlijk weertje.
5
00:03:15,000 --> 00:03:15,005
Ik wil spreken over een karatetoernooi.
Am I right?
Of course, subtitle sequences should (at least) go together with speech, so a fixed duration for each sequence must be wrong.
How to correct this?
Additional speech recognition software may help you a bit, but it surely wouldn't be enough. After all, subtitles should not only follow speech, but also allow enough time to read. Typically, speaking goes faster than reading, so in reality, if the translation text is not shortened (compressed), the subtitle sequence must be visible (much) longer than the actual speech. So, to correct this, you probably have to go through each sequence, one by one, to adjust it. It goes without saying that this is a very time consuming job.