[rugarch:realGARCH] A good proxy for close to close volatility?

Hello,

I am trying to use the realGARCH model, but whatever I tried passing as
realized volatility gives spikes in sigmas here and there (sometimes more
than 1).

My questions are:

1. Can large sigmas be avoided with this model?
2. If daily (close to close) returns are used as the data for the fit, what
can be used as a reasonable estimate for the realized volatility? Does it
make sense to use some volatility estimate based on OHLC data?

Thanks in advance,
Ivan