My unhelpful answer is that we cannot tell.
In the referendum on 22 May, 953,683 people voted - 676,966 Yes, 274,979 No and 1,738 spoilt.
In the Assembly election on 25 June, only 823,565 people voted, which means that there are 130,118 voters (net) who voted in the referendum but cannot be classified as either Unionists or Nationalists because they did not then vote in the Assembly election.
One can break down the Assembly election vote as follows:
185,094 votes for anti-Agreement Unionists (DUP, UKUP, Ulster Independence Movement, Robert Lindsay Mason - UIM though not strictu sensu Unionists probably belong in this category)
321,349 votes for Nationalists (SDLP, SF, Oliver McMullan and Delores Quinn)
201,510 votes for pro-Agreement Unionists (UUP, PUP, UDP)
79,237 for centre parties and other independents
625,135 votes for pro-Agreement parties
It's clear that pro-Agreement Unionists had a narrow lead of 16,416 votes over anti-Agreement Unionists in the Assembly election. However, the pro-Agreement vote was 51,000 lower, and the anti-Agreement vote almost 90,000 lower, in the Assembly election than in the referendum.
Can we map this difference in any more detail? Yes, a little. Although we don't have constituency breakdowns of the "Yes" and "No" votes in the referendum, we do have turnout figures. They show that the turnout differential was much greater in Unionist areas than in Nationalist areas, and that there is an even stronger correlation with the electoral strength of centre ground political parties. So there is some evidence for the view that Nationalist voters turned out for both referendum and election, and voted overwhelmingly "Yes" in the referendum, probably contributing a majority of the "Yes" votes (if anything, I detect signs of some SF voters abstaining in the referendum but them coming out to support the party in the Assembly election).
The rest of it can be interpreted in various ways. Occam's razor would suggest that there was a reservoir of voters, more anti- than pro-Agreement, who came out for the referendum, and then abstained in the election (as they usually do). Do they count as Unionist voters? Not by my reckoning, I think you have to actually vote Unionist to be a Unionist voter. Were they mainly from a Protestant background? Yes, probably.
There is another possible interpretation: that the voters who turned out in May but not in June were actually more pro-Agreement than anti, but that in June a substantial number of UUP voters who had voted "No" in the referendum none the less supported the party in the election, on grounds of traditional loyalty, accepting the popular vote, etc. Occam's razor shaves this one close, but I think it maps better to the subsequent fraying and disintegration of UUP support.
Conversely one could argue that there were vast numbers of voters who turned out to vote "No" and then a large number of "Yes" voters who changed their minds over the following month. But I think that much less likely.
But basically, the proposition that the 1998 referendum was supported or opposed by any kind of majority among Unionist voters is impossible to prove. The numbers can be read to suggest that, as in the Assembly election, a slim majority of those who normally voted for Unionists also voted Yes in the referendum; but the evidence is very weak indeed.