There is a confounding variable (omitted variable bias) here that could explain the entire thing, namely, those who used precise numbers likely ACTUALLY DID use a model and/or more quantitative analysis to determine that price.

Similarly, those who chose hole number prices likely did NOT use a model to determine that exact price (as the odds of a model resulting in a whole number price is extremely unlikely).

Thus, you cannot disentangle the effects of the precise bidders actually being more quantitative from the perception of them being more quantitative by the seller.

To properly disentangle this, you would need an experiment (or potentially quasi-experiment) where you randomize either adding some noise to the bid for some bids and rounding other bids to whole dollar amounts.