Here's a result of the planner application.
https://imgur.com/a/Gyk6hlH
The marked semaphore should obviously belong to another rail, if exit at all.

Interesting, thanks for the report

It shouldn't exists no..
The turn is initially marked as bidirectional if you start with this:
https://i.imgur.com/Lal9EyP.png

It's due to a fix in 1.1 I hadn't tested with bidirectional stuff..
When there are two chain signals in a row, it'll delete the first one.
So during the algorithm this:
https://i.imgur.com/O8tirRX.png
turns into this
https://i.imgur.com/TeenjCj.png
Because at that point it didn't check for bi-directionality

Fixed in 0.1.4