Usando union all
y las subconsultas para limitar los registros deberían hacerlo:
select * from users where id = 5
union all (
select * from users
where score < (select score from users where id = 5)
order by score desc limit 2
)
union all (
select * from users
where score > (select score from users where id = 5)
order by score asc limit 2
)
order by score
Editar:creo que un mejor método es numerar las filas según el puntaje y luego seleccionar las filas con el número -2 y +2 de las filas de id 5:
select id, name, score
from (select
t.*, @rownum1 := @rownum1 + 1 as rank
from users t, (select @rownum1 := 0) r
order by score
) a,
(select rank from (
select t.*,
@rownum := @rownum + 1 as rank
from users t, (select @rownum := 0) r
order by score
) t
where id = 5
) b
where b.rank between a.rank -2 and a.rank+2
order by score;