I agree that a piece in such positions looks better than 3 pawns (which stand on their base for the moment). The 'struggle to a draw' part is there for anyone interested in the Cochrane to evaluate and maybe a couple of other lines down the 5. Nc3 or the 5. d4 c5 6. dxc5 lanes.

Of course the Cochrane is not like playing the QG, it is definitely a riskier approach, but the 'refuted' part may be hard to prove..

I remember Watson commenting that this was definitely favorable for Black but unfortunately he didn't show any variations.

Statistics are indeed statistics, but I would be interested to know the method by which White can establish his alleged draw.

5.d4 c5 6.dxc5 d5 looks hopeless for White, to me anyway.

5.d4 c5 6.Bc4+ d5 7.exd5 b5 and I would not wish to be White.

5.d4 c5 6.Nc3!? looks worthy of investigation, though I suspect that White's comp is not quite enough even then.  6.Nc3 Nc6 7.Bc4+ d5 8.Nxd5 Be6 and here 9.0-0!? is untried in my data base.

After 5......c5 you play 6. Bc4 and then after 6......Be6 7. Qe2 again.  With the same plan except you might need to gambit the d pawn, but i think black is much better off with Be7 in than c5.

In my very brief cochrane gambit career, I believe the only winning chances for white are in the 5. Nc3 line.  Moving the d pawn 2 early or the bishop check will give black lines that he has seen before.
Here is a possible line

1. e4 e5 2.  Nf3 Nf6 3.  Ne5 d6 4. Nf7 Kf7 5.  Nc3 Be7 6. Bc4 Be6 7. Qe2. 

your goal is to keep his king in the center for as long as possible while castling and building up your center with f4.  Pretty simple plan, but hard to execute accurately.

Well, from what I have seen the gambit is not refuted. Now, statistics are ..statistics

I personally prefer the 6. Bc4+ line (i.e. have not extensively researched 6. dxc5) and it is a balanced game. I suspect this could be possible with 6. dxc5 as well (of course some analysis is needed here, which I have not done).

Theoretical status I think is that it leads to ...a draw.

The most critical line imo (and also according to most sources of theory) is: 1. e4 e5 2. Nf3 Nf6 3. Nxe5 d6 4. Nxf7 KxNf7 5. d4 c5!? and if 6. dxc5 d5!?. The idea is to break White's central pawn phalanx by returning the extra pawn.

It is a nice gambit line that is mostly based on long-term positional themes rather than fast attacks and the like. There are some threads here should you require more information (use the 'search' function).

For a summary on sources you can also check this site: http://kenilworthian.blogspot.co.uk/2011/03/cochrane-gambit.html