It is pretty common to sometimes get obviously erroneous HR or power data (e.g. HR = 250 or power = 2000w for someone who has never seen HR >185 or power >850). When looking at past peak performances over any time horizon, this bad data means that you cannot see peak performance data or you have to manually go through and delete HR and power data from those activities respectively before you can see what the athlete's actual peak performances were. This can take a lot of time if the athlete gets bad data a few times every month, and a ton of time if you have many athletes. In principle, if you have a manually established and reasonable upper bound, deleting data outside of those bounds would be totally reliable, because we can be sure that an athlete with known peak HR and power will not suddenly see HR at >110% of max or power at >125% of max, no matter what their training status.