This paper investigates a selection of methods disentangling contributions from price jumps to realized variance. Flat prices (prices sampled consecutively in calendar time with the same value) and no trading (no price observation at sampling points) - both of which are frequently occurring stylized facts in financial highfrequency data sets - can cause a considerable bias in each considered method. Hence, we outline an approach for making these methods more robust so that they can provide undistorted statistical results based on intraday intervals not influenced by flat prices and no trading. The new approach is tested in realistic Monte Carlo experiments and is shown to be extraordinarily robust against varying levels of flat price and no trading bias. Additionally, we examine the new approach empirically with a data set of electricity forward contracts traded on the Nord Pool energy exchange. We obtain coherent conclusions with respect to predefined qualitative indicators.