I think 9s per move is fine, and so is 10s.
Besides the threshold for blitz, I feel that the computation of average move time might be improved a bit. In practice, 9s byoyomi is less average move time than a 9s increment. And for basic time, I feel that the board size should also be taken into account to compute the average move time.
A month ago, I made some proposal to account for different time systems and board sizes: 1s per move ranked games - #69 by gennan
Applying those formulas to 19x19 with Byoyomi, 5m + 1x10s = 10s per move, while 4m + 1x10s = 9.5s per move.
Applying those formulas to 19x19 with Fischer, 4m + 9s = 10s per move, while 3m + 9s = 9.5s per move.