Also, which "Mode" does this quotation refer to?Grain count
Grain count controls the number of grains kept in use by the voice. Although only one grain is played at each moment, the voice can rearrange the grains in time depending on the Mode.
I am also uncertain about "Cross-fade": What exactly does cross-fade crossfade between? (The bands' border frequencies? The grains's overlap in time? Is this a fade-in and fade-out at the start and end of the grains?)
Cross-fade
Cross-fade defines the cross-fade region size. For 0% no cross-fade is used which may result in clicking and glitches.
EDIT: Another question: When MGranularMB slices the input audio, do the slices randomly overlap, or are the slices in sequence with beginnings and ends perfectly aligned?