I have surveyed to evaluate the perception and attitude of consumers towards ultra-processed foods (Sample size-621). I am doing confirmatory factor analysis and Structural Equation Modelling. I have four constructs (Risk perception-3 items, Benefit Perception-3 items, Attitude-a4 items, Behavioural Intention-was 4, 1 deleted now 3) and 24 items. My measurement model fit was good—factor loading from a minimum of 0.44 to a maximum of 0.70. The problem is CR (0.70 for two constructs, 0.59 for the other 2), AVE (ranges from 0.29 to 0.45. HTMT has no issue. I think I can justify it, but I also need advice.
The main problem is I am getting risk perception is positively and significantly correlated with attitude, which is opposite to the theory and other research findings. My hypothesis was that the greater the perceived risk, the more negative attitudes towards ultra-processed foods. In this situation, what should I do?