If you take the first "table" of char sequences and IPA glyphs in
the *Help* buffer from C-h I i p a RET and run it through a pipeline,
"perl  -p -e 's/ +/_/g;s/\|/\\vert{}/g;s/_/  \| /g;s/^/\| /;s/$/ \ |/'" , you'll
have the right data.

Then if you put that into an org-mode buffer and cause the "special symbols" like \vert to be displayed, type C-c C-x \.

Now if you position to the 1st cell in the table and start typing TAB as the cursor moves through the table, I believe you will see strange results.

I'm including a file with the output of the pipeline

Thanks,
Bruce

On Sun, Feb 5, 2017 at 3:38 PM, Eli Zaretskii <eliz@gnu.org> wrote:
> From: bruce robertson <brucer42@gmail.com>
> Date: Sat, 4 Feb 2017 23:33:54 +0000
>
> 1. get empty buffer
> 2. put in org-mode
> 3. enter IPA multilingual text input
> C-x RET C-\ ipa RET
> 4. describe input method
> C-h I
> 4.5 exit "IPA mode" w/ C-\
> 5. take resulting *Help* buffer and put in empty buffer from step 1
> 6. format "KEY SEQUENCE TABLE" as an org-mode table replacing vertical
> bars in data with "\vert{}".
> 7. type TAB somewhere in table.
> 8. The columns do not align.

Can you tell in more detail how did you do step 6?  I tried a simple
replacement with M-%, but didn't see the result you reported, so I
assume that's not what should be done.

Thanks.