I recommend the SEM related books written by the following authors: Hair, Holyle, Kline, Byrne, and so on. There is also explained, what are the appropriate values for Cronbach alpha, etc. I am always applying the tree step procedure for SEM: 1. Preliminary exploratory factor analysis, 2. Confirmatory factor analysis (to obtain the measurement part of SEM model), 3. SEM modeling (to obtain the structural part of SEM model). If the data are only slightly nonnormal, in general it is allowed to use the maximum estimation method. Otherwise, some other estimation method must be used in EFA, CFA, and SEM. Naturally, during the execution of the entire modeling process, the reliability and validity tests must be done, and the GOF indices must be investigated.