Realised volatility from OHLC

Five days of a hypothetical Nifty series:

(a) Compute the annualised close-to-close volatility, once assuming zero mean ($\frac1n\sum r_t^2$) and once with the sample variance (ddof = 1). (b) Compute annualised Parkinson, Garman–Klass and Rogers–Satchell volatilities (mean of the per-day estimators). (c) Compute Yang–Zhang with $n=5$. (d) Which number would you trust most for these five days, and why is the answer to (d) different from "the one with the highest efficiency"?