Expected return on the staking dashboard is calculated using the average of expected return snapshots over the last 30 elected epochs of a validator. This has become misleading since return has stabilized at a 10%-15% level and likely to decrease slightly moving forward. I would suggest we use the data from the past 10/20 epochs.
The most popular complaint is there is a higher chance of missing signatures during network transition (e.g. slot increase) and this surely will happen again in the future thus average uptime is not a good indicator. As Ionut has suggested, the Median uptime over the past 10-20 epochs might be a better measurement.