The current in a bandgap is not temp independent, only the bandgap voltage is.
But as long as the opamp is designed to work within the full current variation range, it's fine.
And from your bandgap, I see I2 and I9 are derived from the same wilson mirror. IR being the sum of the 2, will also share the same temp dependence. In other words, you have no other choice. You either use this, or have to come up with another bias current generator.
And lastly, your bandgap doesn't seem to have a startup circuit.