How do we know the sample has unequally probability sampling, so we need weighted distribution?

I have the claim frequency data from this paper http://erepository.uonbi.ac.ke/bitstream/handle/11295/102825/Makori%2CVanis%20K_Modelling%20of%20Auto%20Insurance%20Claims%20Using%20Discrete%20Probability%20Distributions.pdf?sequence=1&isAllowed=y The data is attached. claim frequency data

I tried to fit the Poisson-Lindley distribution to this data. But, the chi-square test showed that PL Distribution didn't fit for this data. Even though, PL Distribution should fit for the overdispersion data.

Then, I tried to use Weighted Negative Binomial Poisson-Lindley distribution. I found the WNBPL distribution from this paper https://www.preprints.org/manuscript/201805.0026/v1 As a result, the WNBPL distribution fitted this data. As additional information, the Negative Binomial distribution also fitted the data. But, the WNBPL distribution slightly better fitted this data. I still can't understand why did we need add "negative binomial weight function ((x+r-1)!/(x!*(r-1)!)" for PL distribution for fit this data? I think this data has unequal probability sampling, so we need weighted distribution. But, I don't understand why did this data have unequal probability sampling? Advice and feedback would be greatly appreciated!