AAlthough initially labelled conservatively as "A possible start towards a partial answer...", this turns out to actually have been the core of the actual answer, increased to more than 90% with the final comment, so I'll now undelete and "finish off" the part that had been posted as comments for easier access to how that answer was determined.
there's about a constant entropy level within the grid, so a scheme along these lines, even if not divided exactly like that, may be the way to go.
The first (failed but suspiciously close) attempt was to
look at which "vertical" walls were present, and to use the binary code as an index to the alphabet. This resulted in:
01 08 04 B I E 02 15 15 C P P 00 16 16 A Q Q 05 01 05 F B F 14 04 20 O E U 14 24 02 O Y C 11 05 08 L F I 04 01 05 E B F 00 02 18 A C S 05 21 05 F V F 08 08 00 I I A 12 14 00 M O A
It seemed "suspiciously close" because
all indexes were in the range 0-24 despite 0-31 being technically possible. At that point I tried various offsets (e.g. 0=A or 1=A or even 2=A), and got nowhere, so figured there was "something wrong" with the encoding.
I'd meant to look back more closely at the question for hidden clues about what to do next. Instead, "Cont'd 2" was posted, after which it was clear that
the directions of the walls "attached" to each node/pillar were important, so I recorded them:
E E E E N W N E E E E E N E S E E E N W W N N N N W N N N E W W W W S N E E E E N W S W W W S N S W E E S W N E E S W N E E N W E E E E E E N E E S E E N N N S N N W E S W E E E E E N S N E E E N W S E N W W W W W W E E E E E E E E E N W N E E E E S S W S N S N W W N W E S N W E E S W W W E E N W N W S S W W W S S S N E E E E E
This still left open
various ways to extract information from the groups of 5 pillars that still only really remained an "initial hunch" based on the first significant clue of 3 x 11 letters. Even if we're only considering a 5 bit code (which was far from certain at that point, I'd also wondered about other possible ways to have 8 bits encoded in the 10 raw bits available), which two directions should be considered "1" bits and which "0" bits, or is there some other encoding entirely? My first attempt was actually stupidly close, where "E" and "W" are interpreted as 0 bits, and "N" and "S" are interpreted as 1 bits...
01 08 05 B I F 02 15 14 C P O 01 16 20 B Q U 14 05 05 O F F 04 00 18 E A S 15 25 00 P Z A 14 05 08 O F I 00 00 05 A A F 01 23 18 B X S 12 08 05 M I F 12 15 00 M P A
However, I then tried different ways to extract a bit (resulting in different kinds of garbage) rather than trying something that would have easily let me see the actual message:
using 'A'-1 as the base rather than 'A'+0 resulting in:
01 08 05 A H E 02 15 14 B O N 01 16 20 A P T 14 05 05 N E E 04 00 18 D @ R 15 25 00 O Y @ 14 05 08 N E H 00 00 05 @ @ E 01 23 18 A W R 12 08 05 L H E 12 15 00 L O @
Fortunately, @chris was able to see the step that I missed, and complete an answer before @florian-f got even more frustrated at how close I was getting in comments and then just missing the actual answer!
