From my perspective BME280 has the best price to cost ratio. However SHT31 and SHT35 are better in measuring both temperature and relative humidity. Also SHT3x-DIS-F option with PTFE membrane on top of sensor is great for outdoor usage.
In my lab I use two pairs of BMP390L and SHT35 as reference point in test chamber. This is very nice combo, especially with PT1000. Most of original BME280’s were in specs (±0.5°C). Occasionally I saw poorly calibrated units with ±0.7°C. All of them behave in similar fashion following the reference point.
The source you are referring was first published in April 2016, and was updated multiple times since than. It shows many neat charts but you have to acknowledge that Robert (the author) used only one orf two sensor types for each test. Instead of using 5-10 sensors of any given model from different manufacturing batches and purchased from different sources. This is a very nice overview but as data source it isn’t something I would like to rely on.
You also have to acknowledge that for popular sensors there are plenty of counterfeit units on the market, eg: DS18B20, BMP280, AM2301, etc.
Sometimes you are able to spot counterfeits visually, other time by analyzing response times, contents of registers, etc. In case of DS18B20, the great work was done by Chris Petrich (currently working at SINTEF Narvik, previously at Sea Ice Group at the Geophysical Institute at UAF). GitHub - cpetrich/counterfeit_DS18B20: How to tell original from fake DS18B20 temperature sensors. - this example shows how crucial is developing tools for identification of genuine sensors especially in global shortage times.
This is also why I was thinking about manufacturing our own modules from parts ordered directly from manufactures and authorized resellers. We all here because of Citizen Science. I hope this doesn’t turn into Citizen Bullshit Data Source.
Long story short - Yes, I believe that SHT31 is a good sensor