Double distribution in the error of a fitted parameter

[mdgalati]

I am using RooMCstudy to perform 2000 toy fits:

mcstudy = R.RooMCStudy(
        pdf_model,
        vars,
        Binned=True,
        Silence=True,
        Extended=True,
        FitOptions=dict(Save=True, Strategy=2, PrintEvalErrors=0),
    )
    mcstudy.generateAndFit(nMCevents, sum(yields), keepGenData=True)

I am fitting the yields of different components. However I noticed that the error on one parameter is weird, as it presents two peaks. Any idea of why this happens?
(I’ve used two colors to show that the yields don’t show any trend)

[couet]

I think @jonas can help you.