In the following code block, example 10 works as expected. The rg command matches the given PAT pattern. But example 11 doesn’t match anything. The only difference is the b character in the INPUT variable. What is the reason that example 11 doesn’t match?

#  example 10:
INPUT="#f/test"
PAT=$(echo "${INPUT}" | sed 's;/;\/;g')
echo "#f/test line" | rg --vimgrep "${PAT}"
# :1:1:#f/test line

#  example 11:
INPUT="#f/testb"
PAT=$(echo "${INPUT}" | sed 's;/;\/;g')
echo "#f/test line" | rg --vimgrep "${PAT}"
# (none)

In the next two examples, I omitted sed expression that replaces / with /. Now, word boundary b works as expected:

#  example 12:
PAT="#f/test"
echo "#f/test line" | rg --vimgrep "${PAT}"
# :1:1:#f/test line

#  example 13:
PAT="#f/testb"
echo "#f/test line" | rg --vimgrep "${PAT}"
# :1:1:#f/test line

What is the reason that word boundary doesn’t match in example 11?