My feelings on data about feelings

I decided to scrap my initial Instructional Assistant survey. I feel better as I scan over the initial returns from my survey. I am looking for patterns in the data but I am wondering if I am imposing what I want to find on what is actually there? This is clearly potential for flaws in validity of the data. I am wondering how I avoid this. Even in my interviews I feel like I am leading my interlocutor with my questions. My critical question is important, I believe in it, but I am just worrying that I am shaping the results a little to prove something. 

The third data source, observation of staff interactions with students, feels the best to me. When I am observing my staff work with students, I find positive and negative interactions. I am finding that staff are trying to use strategies with students. They only need a little refinement in some cases. More than anything, I am seeing that the staff has goals of their own that they are trying to accomplish with students (finish a project or assignment, get out to job sites on time, etc.) that do not allow the students to learn from the experiences. In this, I see where action research can lead to more action research.