Why do the high quality H.264/AAC/MPEG-TS hour files seem shorter than an hour?

The video backend service records to a MPEG-TS stream and cuts between two MPEG-TS packets at exactly the hour change. This does not take into account if a set of packets belongs to each other. Think of this as word-wrapping in the middle if a word, instead of waiting for a space or a dash to wrap to the next line the word is cut in the middle. When each hour file is analyzed on its own, the interpreting software will probably miss several frames as they are considered incomplete, at the start and the end of the file. This may result into a reported duration shorter than an hour. When these hour files are concatenated together they represent the complete media stream again and nothing is missing.