I have a question.
Is it normal to find a offer/lander combo that pasts the first spend to have potential and then have it fail considerably the next go around after focusing on the converting OS/carrier/landers/offers?
Here is the stats after the first wave of testing:
Attachment 19200
And then I do all the filters to run a 2nd wave of testing:
Attachment 19201
Hey buddy, its not exactly 'normal' but it does happen quite frequently man... although one thing I have to point out is the considerably low spend you're gathering the data from. What I would suggest is upping the budget to spend at least $30-50 and then if you don't see some more stability in conversions on certain " OS/carrier/landers/offers" combinations then move to another offer.
There are many reasons campaigns fluctuate, day of week, time of day, tech issues, traffic source you name it man.
Its just very difficult to draw any conclusions on such a low spend amount, do let us know how it goes if you continue it man interested to see your findings 
But you're on the right path, good job on launching campaigns dude and learning the skill of optimising is a rock solid foundation.
As @mayenzie pointed out, making decisions on small data like this is kinda hard.
You might get lucky in the first round, but not in the second.
Give it a bit more spend to get a clearer picture.
This happens because of 2 main reasons :
1. change in the bidding chain or traffic delivery ... so you're getting different traffic now. Can be a higher bidder taking some of the good traffic away, or a publisher left the network, your campaign was blacklisted by the publisher etc ...
2. funnel wasnt strong enough ... you were able to catch a few eager users early, but some part of the funnel isn't strong enough to perform for a longer time. This happens very often, especially with smaller audiences.
And the small data sample probably plays a role here too, as guys above already mentioned 