Problems with deadlocks in recursive maintenance task

I am having a problem and I want to ask some opinions on the best approach to accomplish this task.

The process that I am implementing is a purge solution for big databases. The tables can have a variable amount of rows that can go from a few thousands to big millions. What the process will do is to delete just the data for a specific year. Partitioning is not an option here.

I have a some initial key tables that were chosen to start the process, and from these tables I will recursively delete data, using some stored procedures built for this purpose (in order to maintain data integrity and delete only the necessary data).

I am using SSIS to be able to start the data deletion in multiple tables at the same time but I am having problems with locks.

The deletion process as the following steps:

1. Drop all indexes on table
2. Delete data in batches
3. Rebuild the primary key indexes (this needs to be done for performance, because there are table with some primary keys non clustered)

I have tried:

1. Using transactions and named transactions (still have locks)
2. Using sp_getapplock (still have locks)

In SSIS I have serializable as the default isolation for the transactions.

My main question is, because I have to do some DML and DDL in the same transaction, and this will take some time to execute, I want all tasks that need to obtain a certain lock to be waiting until the lock is released and not be targeted as deadlock victim (very the same as a mutex).

The database will have no activity during this operation, only this process will be executed.

This is the code I use to delete the data, and here is where the deadlocks are coming from.

select @query = N'

BEGIN TRANSACTION;
DECLARE @result int;                                        
EXEC @result = sp_getapplock @Resource = ''[dbo].'+@dep_tbl_nm+''', @LockMode =    ''Exclusive'', @LockOwner = ''Session'', @LockTimeout = -1;                                      

EXEC CreateIndexes ''dbo.'+@dep_tbl_nm+'''
EXEC CreateCoveringIndexes '''+@tbl_nm+''','''+@dep_tbl_nm+''','''+@dep_tmp_tbl_nm+'''

WHILE(1=1)
BEGIN
   DELETE TOP(50000) FROM a OUTPUT deleted.* into '+@dep_tmp_tbl_nm+' FROM dbo.'+ @dep_tbl_nm + ' AS A INNER JOIN '+@tmp_tbl_nm+' AS B ON ' + @on_list +'

   if(@@ROWCOUNT = 0)                                               
  break;                        
END

exec (''ALTER INDEX ALL ON '+@dep_tbl_nm+' REBUILD WITH (FILLFACTOR = 80)'')                                                        
COMMIT TRANSACTION;'    

print(@query)                                   
exec(@query)    

Comments are welcome

Regards