the maximum ta measurement is not always reliable, because it can always happen that a MS has a wrong implementation and will tend to send its burst too late. Or maybe it’s a little problem in the BTS.
The real question is : how many samples with TA=63 are reported during one day ? If it’s only 1 or 2…it’s not important at all.
The important thing is to check the distribution of samples per timing advance, so you can see where are located 99% of the subscribers.
The difference between los and nlos is not something that we can use in radio optimization, as far as i can see.