I actually think using the website for uploading new data is better than the original competition.

For example, you can see the results of analysis with my data here:

http://cdwscience.blogspot.com/2019/05/precisionfda-and-custom-scripts-for.html

(and the upload format was a little strange - I used this script for re-formatting)

The blog post also explains my concerns with the original competition results.

No matter what, I think there should be some expectation for re-analysis. For example, for the annotation step alone, a 10-20% increase in diagnoses with re-analysis is mentioned in this paper:

https://www.nature.com/articles/s41436-019-0719-3

However, I think doing things like filtering for target regions (and probably filtering out some variants) is likely going to help. Changing the pre-processing and/or variant calling may or may not be necessary.