So without more details its a bit hard to answer this question but in general I would say is that you should always look to use "the right tool for the job" Neo4j is great at answering certain connected data type questions, but not necessarily great if you are trying to scan through the entire DB and perform aggregations across it etc. Neo4j can indeed model time based graphs, with the following blog giving one such example of how to do this: http://neo4j.com/blog/modeling-a-multilevel-index-in-neoj4. Having said that Cassandra also excels at time series data, and there is no reason to not use both if required, depending on what you are trying to achieve! In this polyglot persistence world we now live in, there is generally no one size fits all.
For me the end solution is important, and in my experience this often involves having to use more than one technology to achieve the desired end result.