SPARQL aggregated query with 2 WHERE and 2 GROUP BY?

Question

I am using a database with bands, their albums and the release date of the albums. I'm trying to get a table with one column for the number of albums released in the 1980s, and one column for the number of albums released in the 1990s.

I expected a return table like this:

band	numalbum1980s	numalbum1990s
<iri/ABBA>	12	1
<iri/Chicago	5	3

This is my current attempt:

SELECT ?band (COUNT(?album1980s) as ?numalbum1980s) (COUNT(?album1990s) as ?numalbum1990s)
WHERE {
   ?album1980s :artist ?band .
   ?album1980s :date ?1980sdate
   BIND(YEAR(?1980sdate) as ?1980syear)
   FILTER(?1980syear > 1970 && ?1980syear < 1990)
}
GROUP BY ?band
UNION
WHERE {
   ?album1990s :artist ?band .
   ?album1990s :date ?1990sdate
   BIND(YEAR(?1990sdate) as ?1990syear)
   FILTER(?1990syear > 1990 && ?1990syear < 2000)
}
GROUP BY ?band

Which yielded this:

band	numalbum1980	numalbum1990s
<iri/ABBA>	12	0
<iri/Chicago	5	0

As you see, the query results lack anything from the second WHERE clause entirely. I get a correct count in the numalbum1980s col, but just zeros in the numalbums1990s col.

I also tried encapsulating both WHERE clauses in a single clause, but that query is malformed:

SELECT ?band (COUNT(?album1980s) as ?numalbum1980s) (COUNT(?album1990s) as ?numalbum1990s)
{
WHERE {
   ?album1980s :date ?1980sdate
   BIND(YEAR(?1980sdate) as ?1980syear)
   FILTER(?1980syear > 1970 && ?1980syear < 1990)
}
UNION
WHERE {

   ?album1990s :date ?1990sdate
   BIND(YEAR(?1990sdate) as ?1990syear)
   FILTER(?1990syear > 1990 && ?1990syear < 2000)
}
   ?album1980s :artist ?band .
   ?album1990s :artist ?band .
}
GROUP BY ?band

If you have any thoughts of solutions, I'd love to hear them. (I'm currently using the SPARQL interface in cloud.stardog.com, but I don't think the interface matters.)