I don't want to get in the habit of automatically starting kos. In an actual game starting an un-winnable ko without having useful ko threats elsewhere probably would be "an even greater failure".
How? Let's say that the value of a non-ko-threat move on the board is x. Let's also say that there is a position on this board where you can start a ko. Instead, you play the move of value x. Your opponent then removes the ko aji from the board, enabling you to play another move of approximately size x as well (it could be bigger or smaller depending on whether you were threatening a group, but let's assume that, since your opponent tenukied, your move did not threaten too much). You have now made a profit of approx. 2x not including the ko position.
Now let's say that instead, you start the ko fight. Let's say that if you lose that ko fight, you locally lose y points (if it is a flower-viewing ko for you, as most of the problems posted so far have been, than y is relatively small compared to most other moves on the board). Now, it seems the worst case scenario is that you make 2x profit elsewhere - that is, you are not able to find a single ko threat and so you just get two big moves in a row. However, more likely, you will be able to find a ko threat worth > 2x (even if not drastically so) which your opponent will not respond to. So as long as 2x + y < your ko threat, it was a profitable ko to start.
That being said, it may have been more profitable to wait until the value of your unanswerable ko threat was larger, but I must disagree when you say that starting a ko too early is "an even greater failure" than not starting the ko at all.
Maybe I'll go back to that set in the future if I find myself not being able to start a ko when I need to.
To me, this seems tantamount to a 9k saying "Oh, I'll learn about the L-group and the J-group later when I see it occur in actual play." A couple of problems with this:
1) When you do see it in actual play you are more likely to not know what to do or to screw up because you have not studied it.
2) You are not going to see such a shape in actual play because the shape is not natural to you. I rarely ever noticed the L-group or any other such group before I learned their statuses and how to kill them. It was only after I studied them that I realized "Hey, if I do this and this and this, his shape becomes an L+1 group, which I can then kill!" If I hadn't studied these things until I could see them in my actual play, I would've spent a lot longer struggling to improve to my current level than I did.
Even now, if I've done enough tsumego of a particular pattern, I can drive my opponent to shapes I know I can punish which, had I not done tsumego, I would never have even seen. You may say "Well, that's life and death, and I'm only talking about kos." But they are not separable. For instance, although many people say the J-group is dead, did you know that the person forming the J-group can quite often alter one move in the sequence to make a ko?
Anyway, my overarching point is, you will very rarely see shapes and think "There's a deficiency there," or "There's an unintuitive way to live there" until you've already done problems/studying of similar shapes.
Of course, don't think I'm yelling at you. You can feel free to do tsumego however you want. But if you're doing tsumego to improve, I don't think putting off a particular branch of it for whatever reason is a very good idea.
P.S. Shaddy, I didn't look at your second solution, but the first one is right.
Here is the SGF file, with a couple of comments.
But I suppose that the interactive software requires a full board. Traditionally, problem books often showed a number of problems on different parts of the same board. You were supposed to regard them as independent. Surely for these problems you are supposed to regard them as local, not whole board problems.
and will do the Jie section next. 
