I would do it like this,
Tree Statistics to get the data paths, then cull every other data path itself and then use the Tree Branch component to “list” the remaining data.
The issue with your cull pattern is that it is operating on a data tree so its actually culling the items in each branch of the data tree based on the 0,1 pattern, instead of culling the entire branch aka “grouped data” itself.
Graph Space:
Note how only the odd branches remain (There were 4 groups, now there are only 2):
