@Ottavs First of all, it's in knots, not mph. Second, sustained wind measurements are much more complicated:
https://en.wikipedia.org/wiki/Maximum_sustained_wind

Basically, it's a 1 minute measurement. That's it. Windy is showing you the average wind speed for an hour. They could show you only the max wind speeds everywhere for the hour, but it would likely be very misleading.

@Ottavs It's only measuring how much is dropping for that hour, it's not adding up all the rain over time to a single area. 1.2 inches of rain in an hour is a lot.

@Ottavs Right. It's showing averages. You could have 3 measurements for a city, for example, 50mph, 50mph, and a gust of 150mph. The average is 83mph.

@Ottavs Because it is showing averages, not peak measurements.

@Ricky_Lightning See my post, this is not a "data withholding" problem. It's math. There are only so many ways to aggregate a bunch of measurements into one while preserving the character of the data.

@Siff The explanation makes perfect sense from a data science perspective. The grid is aggregating data...it is a window function. The simplest example is an average (rectangular window function). Take the average of 10, 10, 20, & 80. The average is 30. Clearly, 80 was included, but the output (30). When the OP says that it "smoothes" away extremal data, they literally mean smooth:
https://en.wikipedia.org/wiki/Smoothing

That said, there are better methods that can preserve extremal measurements. These are called "High Resolution Schemes" or "Total Variation Diminishing" methods:
https://en.wikipedia.org/wiki/High-resolution_scheme
https://en.wikipedia.org/wiki/Total_variation_denoising

These let you recover data like this:

The challenge with using these for weather is scale and computing power.

Stay safe out there!