Regarding the unexpected values, you may wish to perform some QA/QC checks on your data to determine if you have misclassified measurements. Try the Locate Outliers tool or the LAS Point Statistics as Raster tool with the Z Range option to get a sense for whether you have extraneous values. If you manage to capture them using Locate Outliers, then you can optionally choose to reclassify those points using Set LAS Class Codes Using Features.
Regarding your second question, the surface gets regularly sampled with points over the area defined by the polygon. The sampled points are interpolated using the method selected in the tool, then used to create a surface that matches the polygon's boundary. If you don't specify a sampling distance, the tool uses the densification of the LAS data as its sampling distance.