The right approach for you depends on two major factors:
- Why are you recording your sources?
- What are the strengths and limitations of the software you use?
The main reasons for recording sources (and all their details) are (in general):
- To enable you to revisit a source, to refresh your memory about a detail (e.g. an age in a marriage certificate that doesn't tie up with the age in a death certificate) or to understand why you reached a particular conclusion based upon a particular source (his army enlistment papers said that he was 18 when he enlisted, so he must have been born sometime between X and Y). [You'll also want to avoid paying twice for the same certificate, or newspaper image or whatever... And there's always that eureka moment when you realise the witness at your great-grandfather's wedding was his step-father so now you know what happened to his mother after she was widowed... ]
- To help you assess the reliability of a source, or compare the reliability of two sources for the same piece of information, especially when they conflict -- an official death certificate is more likely to be accurate than the memory of Aunt Ethel that her grandfather died when she was two years old. If all you have is Aunt Ethel's memory, you might want to look for another source; but if you have the death certificate you most likely don't need another source for date of death. And if Aunt Ethel remembers that granddad died in 1945 and the death certificate says 1947, you can decide which to believe.
- To allow others to consult the sources you've used, to assess/confirm the accuracy of your research (you might be wrong -- I have been, often, and very grateful when it was pointed out; ask me sometime about Samuel Brooks versus Samuel Wood Brookes, both born circa 1837 in Nottinghamshire 15 miles apart) or to use those same sources for their own research.
If you want to help other consult your sources, you'll provide as much detail as you can, and record each source (document) separately. If you only want to support your own research and enable yourself to refer back, fewer sources may suit your purpose.
Whatever your reason for recording sources, you probably want to record the specific location of each piece of data you've relied on, and to record the details that affect the reliability of that data. ("Aunt Ethel said... but quite frankly, she was a shilling short of a pound note at the time" or "The death certificate said he was 70, but the informant was a neighbour who'd only known him for 9 months").
Then the strengths and limitations of your software come into play.
- How easy is it to create a 'source record' and/or a 'citation'? Is it easy to create a source 'reference list entry', with a full description of how to find it and what's in it (including a transcription), and to link multiple 'facts' to that source (with a 'where within source' qualifier if necessary)?
- How easy it it to create multiple citations to the same source (as very few sources support only one piece of information)? How easy is it to copy a citation when you discover you need to do so, and just change the 'where within source' detail?
- Where can you link an image of a source -- to the source record, to a citation or to both?
Most modern software package won't balk at many thousands of 'source records' or many thousands of citations. And creating a single source record, or single citation, is probably the same amount of effort.
The difference in effort will come if you need to refer to the same source more than once. A UK birth certificate will support the name of an individual, their birth date and place, their father's name and occupation, their mother's name and maiden name, the name and residence of the information which is usually the father or the mother... You'll need to create a citation for each fact; or one source linked to each fact.
My own approach is a hybrid.
I create a separate 'source record' for every unique document (say, certificate) that I consult, and link an image of that document to the source, and include a transcription or abstract (depending on the nature of the document). Then I link every 'fact' for which that document provides evidence to the 'source record' with a 'where within source' qualifier to identify exactly what piece of information I relied in.
But if, for example, I've only consulted an index of documents rather than the document itself, I'll use a general source and a specific citation. So, Aunt Alice's death year/quarter can be determined from the UK GRO Indices (she had a particularly unique married surname and lived to be 110, so I'm pretty sure I've found her in the indices, and there's no information in a standard death certificate that I need for other purposes) therefore I've recorded the source as "GRO Death Indices" with a repository of the GRO. 'Where within source' gives year/quarter and registration district, plus her name and age -- enough to enable somebody else to look her up in the indices and order the certificate of they want to spend the money. (If I do decide to pay for a certificate, I'll include the GRO index reference as a 'source note' to help others order the same document.)
If you do decide to go the route of a very small number of sources (e.g. certificates), I would still suggest splitting them out by 'issuing authority' (state or county) and 'certificate type' (birth, marriage, death, ...). Otherwise you'll just swap a large number of sources for a single source with a large number of citations each with a very long name.