Yes, my current working theory is that it starts with the seed and then runs through a loop and performs various algebraic expressions on the data as it reads it. It may be similar to a CRC calculation, but not exactly like it.

The seed that we have was derived from empirical tests (somewhat a long process, but we've been able to decode signatures up to 17 bytes... almost 18 bytes). A good friend of mine who is an experienced software archiecture is helping me--when we were initialize analyzing the initial data he was decoding it quick that I could capture test cases. At any rate, since the seed that we're using were calculated from the raw empirical data, it might not be the actual number in the code. There are likely other initializations that occur.

It doesn't suprise me that 01 00 00 00 ... 08 doesn't have the same signature as 08 00 00 .. 01.... depends on exactly how the calculations are done.