Optical alarms are better than ionising alarms for -most- applications. Optical alarms do not do well with smoldering fires, as the wafting smoke does not produce enough obscuration/scatter for long enough such that the alarm will actually reliably trigger. However, on almost every other test, they are superior. Therefore the general recommendation is to use optical or hybrid optical/ionisation sensors.
It takes around 0.1dB/m of smoke to trigger an alarm. Obscuration of 3dB/m is half light attenuated transmitted per metre. 0.1dB/m doesn't sound like much but it's a lot once it reaches the sensor; the room has to be seriously smoky. Most detector designs don't trip below 0.05dB/m because the stability of the sensor (especially when you get people smashing into them and knocking them off the ceiling) over 10 years is hard to guarantee, and then dust and other things can set them off. Note most optical sensors work on a scatter principle instead of obscuration, as it tends to be easier to design and stabilise, but the behaviour is similar.
The relevant standards are fairly strict about false alarms and alarming too soon, as they tend to be the detectors that get covered up or have their batteries removed. It is judged to be better to be slightly less sensitive than over-sensitive, as with a real fire the devices will still trip well even if they are 20% less sensitive, but they will not trip at all with the batteries removed.
You aren't wrong about the ionising detectors being a pain, our consultant supplier for test tunnels doesn't like to deal with them because they invoke another tonne of paperwork regarding radiation and storage of components. But you can still buy cheap ones. In mass production they are cheaper to make. Optical sensors require discrete components in the smoke chamber, as well as a photodiode amplifier, usually some software to compensate for drift and the battery performance is a bit worse as the sensor is active and not passive.