And you can still compensate for interrupt latency by subtracting the current timer value from the reload value - so you get a non-multiple-of-32(256) 16-bit divider for your timing ticks (but the interrupt service routine gets a bit more complicated). Unfortunately this technique is rather part specific and it may be necessary to have other interrupts running so I would prefer not to use it.

If you can not, for whatever reason, use T2 (or another reloading timer in your deriavtive) then you have one option for simple timing:
select a crystal that makes your counter value such that the start value of the TL is 0. That way, you cam reload TH only.

The "right way"
use T2 or a PCA

Erik