Correlogram exhibits ACF cyclical - ARIMA Modelling

[bschroder]

Guys,
I'm trying to use Box Jenkins methodology to fit an ARIMA model (with seasonality) to my series. After taking first difference, the ACF correlogram exhibits cyclical behavior which does not vanish with time. And PACF has significant spikes in different (and unusual) lags.
My data is daily.
Does anyone has any idea on how to model such an awkward time series?
I've attached my data (see Plan5).
Thank you