My recommendation would be to observe at least 20 minutes per point, preferably 40 minutes. You need to have enough to ensure a solution where 4-5 minutes in my opinion isn’t enough.
Float can be up to half a meter off, especially in height.
If possible, do the survey when there are no leaves on the trees, leaves will kill your signal even further!
I can also recommend multiple bases when processing, and then processing them together in Least Square adjustment network, using i.e. EzSurv. But time and money wise you are approaching a Total station task.
Check this project I did in quite challenging conditions: Using 4 x RS2's for precise results in challenging environment