You will theoretically require one receive location for all Xml documents. The Xml Disassembler is able to derive the message type from the document.
Then one receive location for each different flat file or custom format. The Flat File disassembler requires the message type be predefined as flat files have no standard mechanism for defining message type like Xml with its root node and target namespace.
One possibility is a custom disassembler that can determine the message type and call the standard Flat File disassembler
If you would create this then you could have one receive location with two disassemblers in the pipeline. The Disassembler stage in a pipeline is different from the other stages it uses FirstMatch execution mode. It calls IProbeMessage.Probe for each component until a component returns true.
The other factor determining how many receive locations you have is the adapter used. How are all the suppliers sending messages? What transport protocol are they using?
I would still go for multiple receive locations, less code to write and maintain, you also get all the tracking and debugging facilities of Biztalk. It will also be easier to bring on new suppliers, just add another receive location, schema and map. With a manual disassembler every new supplier will be a new version with complete regression testing