The Number is the unique key for the data.
Having written this problem down, and examined it as I try to explain it :), I have decided to attempt this approach:

1. Remove all blank lines.
2. Find the index and grab 70 lines (-2..68).
3. Split the data into three sections.
4. Deal with section overlaps.

The three sections are:

1. All lines up to (but excluding) the first line with a colon.
2. All lines with a colon.
3. The rest.

Count 'The rest' and move that many lines from section 2 into (preceding) section 3.

This should help sort the data.

<a href="http://www.graq.co.uk">Graq</a>