The change sets (CB) are not consistent with the ICs

[GreenfishK]

I found that most change sets (added triples + deleted triples) do not fit to the corresponding ICs.
Example for hourly BEAR-B dataset:

IC1 has 33.502 triples, as stated in the paper.
IC2 has 33.502 triples, so meaning added and deleted triples should be equal. However:
Number of added triples in version 2: 155
Number of deleted triples in version 2: 81
Difference: 155 - 81 = 74
33.502 + 74 != 33.502

This is the case for many other changesets, too.
Do you know what might be the reason?

[rubensworks]

Never encountered that myself tbh.

Could either be a bug in OSTRICH, incorrect ingestion on the user's side, or a mistake in the dataset description.

[GreenfishK]

I computed the triple counts and differences for all ICs and CBs. Check it out my notebook in my OSTRICH-BEAR fork.
The main BEAR author is seemingly not available anymore, that's why I posted this issue here with the hope that you might have noticed it.

[rubensworks]

Perhaps @opelgrin knows, he worked with it recently as well.

[opelgrin]

I did my own checks, using the GNU diff tool, and I also find that the changesets for BEAR-B don't match the IC.
It also seems to be the case for BEAR-A.
I didn't check BEAR-C.

Now I don't know which of the IC or CB files are correct and which are not.