The clustering service will always be called, but you can make it so that the matching service is not called until email is populated by customizing clustering so that you only generate keys if all your required fields (first name, last name and email) are populated. This is because the matching service in EDQ will only be called if the driving record has at least one candidate record that it may match against, where the candidate records are selected using the logic that any record that shares any key value with the driving record is a candidate.
Note that with 'typical' (level 2) clustering of individuals, this will actually be the effect, unless you are using other secondary identifiers such as phone numbers, addresses etc. (in which case, are you sure you don't want to use them in matching?) If you do not want to use these a simple way of disabling them would be not to map them from Siebel in the DQ field mappings. EDQ-CDS will only return a cluster key value if all the attributes that form part of the key value are populated, and with level 2 clustering there are no keys based on just the name fields.
If you mean email actually gets a string value of 'NULL' that is being passed to EDQ, I believe this will have no effect since I think the services will only generate keys based on email, and use the email in matching, if the email is syntactically correct (by regex), but you may want to check this behaviour using the Web Service Tester.