what u already tried should have worked.
if u r using replication manager then
make job as broken(should be same as job shown in v$lock).then run the job again.
that should release the lock.
in windows NT u can use orakill sid thread to kill a job.

go to command prompt and type orakill.that will guide u thru.
Here for sid u have to use the sid for the instance(like orcl) and not the sid from v$session.
let us know.