Hi,
1. Acceptable p-value and walds depend on the data, data quality, the type of variable (dummy categorial vars can have high p-values) and what you are trying to predict. Models with a high general p-value should not be used.

2. To test interactions in models, you can use software that has what if? scenario and sensitivity tables. These tools give you the model outcome for a set of values (what if scenario) and the outcome for a range of values ( sensitivity tables). You can try and use a software named Analysis Studio, it doesn't have all statistical measurements and options but it does the work and has a good free license: http://www.appricon.com/index.php/downloads.html