1. Now when I do a combination puzzle (for pattern recognition based training), I get about a roughly 95% accuracy rate, and I never guess the solution before I do the first move.  They average me 20 seconds to solve (and less time to guess the winning move).  But when I did simpler problems, I was told if a question took me over 10 seconds to solve, it was probably too hard.  Does this mean quite simply, the questions I'm doing are too hard?  Or is this increase in time acceptable considering the added complexity of the problems?
   
2. I believe in doing diffucult puzzles without time restraints, like the chess.com daily puzzle on occasion.  I believe this helps me to practice systematically assessing a position, rather then the pattern recognition of time controlled questions.  Does this deserve any of my time?  Would solitare chess (Which would incoorperatate both analyzing a position & calculating it) be a suitable alternative?
   
3. In fritz, there is a training feature where you find loose pieces, possible checks, and attacked pieces.  I'm slow at noticing these things OTB so I think it wise to spend some time practicing these components. Do you think this deserves any of my time?

I see nothing wrong with having to spend 20 seconds on a puzzle. If you're still getting them correct, I would think that these would be at least as effective. Besides, you'll start getting them in 10 seconds after a while anyway.

Also, I think both types of puzzles are worth your time. While harder puzzles without time restraints are good for increasing playing strength, time controlled questions help you find moves quicker so that you can find them OTB in actual tournament play without losing on time.

If you are a bit slow finding hanging pieces, checks, and attacked pieces, I would think that it would be worth it to work on it. You never know when it might come up in OTB play...

thanks for the input, but easier puzzles SHOULD be done in 10 seconds for sure, if you are doing puzzles that take too long you can end up spending more time checking variations then actually learning.  I'm sure 20 seconds isn't killing me however...

3 secs is my time  limit 1 sec is normal if i don't see somthing i probaly will get it wrong anyways

I doubt there is any training difference between 10s and 20s. both work on developing pattern recognition and, given 95% accuracy, efficient basic blunder checking automation.

slow, complicated problems complement perfectly the lack of calculation training of the fast problems.

but It sure is encouraging to see 40 right answers and 2 wrong ones ^_^.

However, I overtrained doing these tactical exercises so much when I actually went to a g/60 tourney I overexerted myself...  I was winning the entire game against a B class player, but resigned because I wasn't able to play anymore ;_;