The results are wrong when combining a pull dimension (when using a convolved as awh-potential) with an FEP dimension (always using an umbrella potential). This should be disabled for now.