How to limit COUNT(*) for SPARQL - sparql

Question

I have a query like below. My goal is to count, for each item that itemX is related to through a specific relationship, count how many has the specific relationship with the item.
SELECT ?b ?c (COUNT(*) as ?count)
WHERE{
itemX ?b ?c.
?a ?b ?c.
}
GROUP BY ?b ?c
This is a hypothetical result
?b|?c|?count
rel1|item1|100
instanceof|human|10000000
rel2|item2|1
I run this query through the Wikidata (with appropriate modification.) The issue is ?count can be at millions and easily hit the 60 seconds time limit.
Therefore, I wish a compromise. I still wish a count, but for each ?b-?c pair, I will only look for maximum for 10000 ?a. For instance, if ?b is instance of and ?c is human, there are millions of humans and listing them all before aggregating may result a timeout. Hence, I wish the query will just list only 10000 ?a for instance of and human and then aggregate it. The result is below is the desired outcome.
?b|?c|?count
rel1|item1|100
instanceof|human|10000
rel2|item2|1
I tried something like this.
SELECT ?b ?c (COUNT(*) as ?count)
WHERE{
itemX ?b ?c.
{
SELECT ?a
{
?a ?b ?c.
}
LIMIT 10000
}
}
GROUP BY ?b ?c
However, I get this result instead.
?b|?c|?count
rel1|item1|10000
instanceof|human|10000
rel2|item2|10000
I wonder where I went wrong.