Oh, I see. Thanks!

Description says:

If a character provided is not in the opposing alphabet, simply leave it as be.

Your solution seems to remove it.

I'm getting a failure for the last test case I can't access until I solve the problem:

"Expected: t_fo&*83, instead got: tfo"

I think we're missing some indication, and probably an example involving the rules implicitely tested to be added to the instructions.

I concur. This is an horrible exercise, one should never push people at mutating references that way. The function should return the list transformed. If you really want the partition point index, you could return an array of arrays, as any decent implementation does elsewhere.