In general, the BandPass filter is used to isolate one type of 'ess'. The BandPass filter 'looks' at a narrow range of frequencies. The 'ess' sound in "Sweat" carries most of it's energy around the 6kHz mark (male voice).
In "Sugar" most energy lies around the 4kHz mark (also male voice). Use the BandPass filter if you only want to reduce one 'ess' and not the other.

If you have several different types of 'ess' noises you wish to reduce, the HighPass filter is recommended. DeEsser responds better in HighPass mode when applied to, for example, a vocal track containing a variety of 'ess' frequencies and vocal inconsistencies. NOTE: the HighPass filter increases the chance of compressing other high frequency sounds such as strings and high pitched instruments.

Most compressors containing a sidechain component can be used as de-essers. However, they require an external equalizer to select the frequency range for reducing. Such compressors have Wideband audio passing thru them and thus compress all frequencies equally.
This is the most traditional approach for purist or vocal-only applications. For powerful 'esses' and the more modern vocal styles, splitting the audio into 2 pieces is more effective and results in fewer side effects on frequencies outside the 'ess' range. If you must de-ess a mixed track of instruments and vocals, the Split method is nearly essential.
Choose the Split mode for the most accurate and strong de-essing and for mixed tracks.
Choose Wideband mode for gentle voice-track-only de-essing and when you do not want to have a crossover in the processing path.

Only to make it easier to adjust the Freq. You can "dial in" the best Freq value by simply monitoring the SideChain. Simply listen to the sound as you adjust the Freq until you believe that you are hearing what you want to de-ess and as little as possible of the remaining signal (that is, not too much of the voice itself).