The short answer is that you cannot create a "correct" citation because you have nothing to cite.
The purpose of a citation is to allow another researcher to identify, locate and re-examine the evidence that you have used. Obviously no-one else will be able to present on the exact car journey where you had the conversation (which is now in the past). The best that you can do is to create a (derivative) source that can be cited by setting out your recollection of the conversation in the form of a document. Ideally, you would send your version of the recount to the other party to the conversation to have him or her confirm that it is a correct record.
In GRAMPS, I would then create a new Repository (called Personal Collection) and enter a new Source (Recount of a conversation ...).
You can use (one or more) Source Notes to set out the details of who you spoke to, when the conversation took place, when and how you constructed the recount, and how the other party corroborated your view to form the final document. Since you have the document electronically, you can upload a pdf image into the gallery with a link to this Source.
Then for each event about which the (now documented) interview provides evidence, you would create a new Citation that points to a particular element of that source. You might use paragraph numbers or even line numbers in the document for specific references.
If the document is complex, you might consider inserting bookmarks within the pdf and using these where GRAMPS specifies volume/page etc.
GRAMPS will then tie all these pieces together when you export a GEDCOM as shown in this fragment.
0 @S0081@ SOUR
1 TITL Recount of Interview with A N Other
1 AUTH My Self
1 PUBL unpub
1 ABBR Aunt Annie
1 REPO @R0021@
1 NOTE @N0266@
1 NOTE @N0267@
2 DATE 30 NOV 2013
3 TIME 17:09:35