Some beginners questions

[ilsebaumgartner]

I am a beginner user of smartPLS (and a beginner in PLS in general). I have done a survey (resulting in 105 responses) regarding the intention to use a specific IT product. The theoretical model is very simple - there are 12 reflective constructs which I measure using 3 items for each construct, all these 12 constructs are hypothesised to directly influence the intention to use construct (which is measured using 2 items). I have constructed the model in smartPLS, assigned all measurement items to their constructs and run the calculation (using factor weighting, data metric mean 0, var 1). The item loadings are fine (all between 0.71 and 0.91), AVE, ICR and Cronbach's alpha within the recommended levels. The path coefficients (I suppose to determine them I have to look at the model and the figures above the arrows connecting the 12 constructs with the intention to use construct) are low - reaching from 0.02 to max 2.5.
There are two issues which I do not understand. Firstly, the R square of the intention to use construct seems to be extremly high (0,944109). And secondly, there are extreme high latent variable correlations (and the sqrt AVE of each construct is in almost all cases lower then the correlation between this construct and other constructs in the model). I would ask for some hints what I might be doing wrongly.