select distinct slower in 5.0 than 4.1

[Saar Barhoom]

I have the following very simple statement:
select distinct(PID) from lastnames where name = 'Smith'

The table has only two columns, PID and name. It has a BTREE index on the name column, and a unique contraint on the PID, name columns combination. The storage engine is MYISAM.

On 4.1.16 this query runs very fast. Using the EXPLAIN SELECT I see that the 'type' of the query is 'ref'.
On 5.0.16 this query runs very slow (about x20 times slower). EXPLAIN SELECT show that the 'type' of the query is 'range'.

Could that be related to the loose index scan feature of 5.0?
Have I forgotten anything when doing the migration?

Thanks,
Saar