ARIMA(0,0,0) model but residuals not white noise

I have a dataset where I am trying to fit an ARIMA model to a stock return – the data set is stationary. I have used the Auto.Arima function to select appropriate AR and MA terms, and BIC selects ARIMA(0,0,0) – which indicates white noise.

However, when I conduct a Ljung-Box test on the residuals i cannot confirm white noise.
Can a conclusion to this be that even though BIC recommends ARIMA(0,0,0) the LB test suggests that it might not be white noise after all?