Welcome, @XavierB
Your observation is interesting, and I’m glad you solved this!
I’m wondering if there is a small bug with the latest wrapper, or if the tools are actually not compatible, or only incompatible with certain use cases. I’ve open an issue ticket with the developers here. → Reported bug: ambertools_acpype/21.10+galaxy0 not accepting gromacs inputs · Issue #180 · galaxycomputationalchemistry/galaxy-tools-compchem · GitHub
They will probably need an example to confirm the problem. You can post a Galaxy history share link to the issue ticket, or back here in this topic. Simple is better – a history with just the inputs and outputs (from both versions) is probably enough.
Thanks! 