Smart List filters will pull names from both Marketo and SFDC. If you're using SFDC specific filters, you then the lead must exist in SFDC to have this data, i.e. Record Type ID or Lead Owner information
I've had to create a custom field to help identify which leads exist in Marketo only and which are in both Marketo and SFDC. Leads that are in SFDC only will sync to Marketo and create a lead in Marketo automatically, so the best way to think about it is that Marketo has visibility into the entire Marketo and SFDC databases, and therefore SFDC does not have the ability see Marketo only leads.
Seeing as you have leads that exist in Marketo only, I'm assuming both databases aren't entirely synced. If it is completely synced then every lead/contact would exist in both databases.
Does this answer your question?