From what I gathered, storm is stateless and an external system is needed to have the messages persisted while the data gets processed. This persistent piece is Kafka, where external sources push their data to. On the other hand, flume is designed to stream data, where it receives data on the source and pushes it to the target endpoint via its attached sink. This is different from Kafka where another system will pull from it. Based on that, flume may not be a good substitute for kafka within the storm and Kafka use case. Although, there seems to be some information on the web about a kafka sink for flume, in which Flume can collect data from multiple sources and deposit it into a Kafka cluster that will eventually be picked up by Storm. As of this time, kafka sink is not yet part of the HDP distro or Apache for flume.
Hope the information helps.