Blake/Runes combined book code

My thinking about the matrix below was designed to address the fact that the symmetry in the matrix reduces the utility of ordering its component cells to find a key to anything. So I addressed that by assembling a book code as follows: Numbers in the matrix used to pick characters from Blake's Urizen
(link at http://ebooks.adelaide.edu.au/b/blake/william/urizen/) I simply counted characters, excluding spaces and punctuation. Two digit numbers come from the preludium. Three digit numbers from Chapter I verses 1-3. Rune words in the matrix are replaced from rune characters in the Liber Primus, counting runes by page/character with 'A Warning' as page 0, 'Some Wisdom' as page 1, etc - I did not use page/line/character because there are several numbers in the matrix that end in zero, which could not be converted to runes.

Use the matrix below for reference.
272             138             shadows         131             151
aethereal       buffers         void            carnal          18
226             obscura         form            245             mobius
18              analog          void            mournful        aethereal
151             131             cabal           138             272

We get:
K	R		IA/IO	N		N
L	[11 dots]	R	J		I
[,]	I		AE	Y		M
I	J		R	[11 dots]	L
N	N		IA/IO	R		K

So drop cells with non-letter characters; we get
K	R	IA/IO	N	N
L	---	R	J	I
---	I	AE	Y	M
I	J	R	---	L
N	N	IA/IO	R	K

Now, we have a non-symmetrical matrix. We also have a matrix with most of the letters of onion in the first and last rows, so a column transposition is a reasonable possibility here. But there are still 3 too many characters here for an onion address, so I think this is a dead end and that I was just seeing patterns.