Even though the message was homophonic, we did of course use frequency analysis to start with. As we had suspected, this did not give us any good clues. It however became clear that the frequencies in the message exhibited large variations.
The presence of the character * in the message made us feel that there was some trick involving that character. There is also an abundance of the letter X. We quickly developed the theory that each digram in the message coded a plaintext character. However as this gave us nothing, we quickly expanded the possible complexity of the coding scheme.
We now worked a while on the assumption that some cryptotext characters encoded only one plaintext character but that some characters in the cryptotext where magic and always started a digram.
Working along this line of reasoning, Staffan wrote yet another interactive replacement program. This supported interactive replacement of either single characters or digrams into the text. It also showed the frequencies for the most common characters and digraphs as one worked through the message.
Using this program, we spent a number of hours trying to get the message to yield, all to no avail. When explaining the way of reasoning to someone else, he immediately replied that it would be better to assume that the encryption was a little bit simpler than we thought. We were assuming the encryption to be almost diabolical as we invented one possibility after another.
The opening to the whole decryption came when Staffan and Fredrik looked at the message and realized that the distribution of some characters in the messages was very interesting. It turned out that the cryptotext characters M, G, C, Z, N and B appear in a very interesting pattern in the message: Each one of them is only present on a few lines and is very frequent in these lines. This made us assume that they all encoded the same plaintext character.