Builty data

I'm currently working on Builty building permit data and I have a couple questions:

I understand that the way the data is collected is: by scrapping over building permit data released by local governments (cities, states, counties) on their website. Is that right? Or do you also collect this data in another way, for instance by looking at paper records?
Relatedly: do you take the information as it is in the original data, or do you process it? And do you take all the columns from the original data or just some columns?
Why is the DESCRIPTION column having missing values, is it because of differences in the span of the data collected across local governments? Or is it for another reason?
Same question for the JOB_VALUE column.
Do you know how to distinguish between building permits that are related to construction work done for building a new building, or related to modifications on an existing building?