parquet gives different count for queries that are similar to each other

Question

I want to read data from parquet from first of Oct 2020 to end of Dec 2020. But with two approaches, I have different results. Here is approach 1:

%spark2.pyspark