I've got a method and theory, and I want to know what you think of it.
The Keyword is irrelevant. The gave us our clue on how to solve it by leaving out null letters to even out the columns, therefore we can conclude that they the last two columns should in fact be the first two (with 2 combinations). That leaves us with the remaining 4. This can be arranged in 4! possible ways. (24 combinations) Combine that with the two and we have a possible 48 combinations for this code.
After converting into columns we can use any polybius square (mine is just a-z plain and simple) to convert it into letters.
We can then use generic decryption methods (like the ones grandmas use when solving the crptoquote in the newspaper. Unfortunately we won't have the benefit of spacing.
The only problem with this is that only 1 solution will work and 47 will not that will leave us a lot of time looking for something when there is nothing to be seen.
For example I reorder the original columns from 123456 to 561234 giving me:
Using my generic polybius square I got
rnu ssl tgn
nzo mso msn
mta ook pps
I did my best to decrypt it from here and all I can get is
"to seen chood prepre or cappille buoncy"
"nsfee boys sp the the shout tilled as bor"
So I don't think this is the one, but I am determined to keep trying.
So does that mean we could try and use that 4x4 grid and place it within the 5x5? I read somewhere that the code in the square doesn't have to be written horizontally or vertically, it can be written in any order. upon making it that is.