Is There A Way To Do A Partial Distinct? (very Complexe Problem)

I'll try to explain the situation as simply as I can.

 A master table exists that has several columns which can either contain trade IDs, or a joint ID of two trade IDs.  These trade IDs exist for contracts.  Contracts can contain many trades (e.g. Contract ABC is written for 40 trades).

 There exists a need for a list of documentation for these trades/contracts that must be taken from the master table, based off the trade ID, and if the documentation type matches a certain criteria.  Because the users cannot understand a simple list of document IDs, added information about each document must be taken from trade/contract tables about each piece of documentation available.  The problem is, two different types of documents can exist for the same contract, but we do NOT need documents for each trade in the contract.

Currently, the stored procedure that someone else wrote is incredibly inefficient and regularly times the server out. 

It begins by creating two identical tables (called "first" and "second" from here on) set up to hold all relevant info about these documents. 
It then selects into one table document IDs for all trades that can be found in all of the relevant columns in the master table, including trades found in "joint IDs." 
It then creates a cursor and selects each row, searching several different things based off the trade ID and updates the first table with these new values.  This is the HUGE hangup because there are full table scans on a handful of tables about 4 times for every row in the table.
Then, after it updates all the information in the first table, it goes about weeding out the duplicate entries by doing the following:
Creates another cursor and does a distinct select on 4 columns which creates a unique ID to pull back exactly the right rows per contract per document from the first table. 
It then stores the information for all of the unique documents by inserting into the second table doing a "select top 1 * from firstTable where column1+column2+column3+column4 = 4DistinctColumnsFromFirstTable." 

 This, as you can imagine creates another huge sink by inducing table scans from a table resident only in memory... AGAIN for every row.

It finishes by selecting all rows from the second table.

Now... what I've done so far, is I've managed to create a complex system of joins to grab all relevant information in one select (steps 1-4).  I have not stored this information anywhere, nor have I used any cursors to do so (obviously as it is one select), but I imagine that I'll have to store all of this data and do a select on the stored data.

 

My question that would solve all of this is, can I do a partial distinct select?  As in, can I select 7 columns where 4 of them are distinct without having to go through inserting into temp tables, using cursors, using top 1... etc.  If not, is there some more efficient way of selecting out an entire batch of info based on only a few columns?

 

If you made it through that mess and decide to help me out, I sincerely appreciate your time....