When we want to prove the presence, or absence of single elements to another party in a secure and bandwidth efficient way, instead of sending whole bloom filters
As stated in the paper several times, bloom filters do not prove presence of an element in a set, only absence. Which is fine when used in a caching setting or where false positives are acceptable, but it seems like if you are relying on this exchange for "proving presence or absence...in a secure way..." false positives are a deal breaker especially seeing as how they are doing "the block chains"
But there are supposedly more papers in the pipeline that might explain this. Either way, it was good advertisement, I went to their site, and looked for the project on github (did not find them though) maybe they could link it from their site?