The unreliability in inherent with the protocol... not anything hardware based..
X10 was designed in the 70's... it was state of the art back then...
and even with a light sensor, it's a band-aid that will cost more...
I'm trying to figure out HOW you would add a light sensor, without adding MORE cost to the idea..
(and still come in cheaper than an ICON module)
You'd need a light sensor for each X10 device.. then an interface for said lighting sensors...
If you send a command to each lamp, one at a time, and wait for the light sensor to confirm that the command was executed by comparing the command's expected effect (ie, the room got a little darker, or it got a little lighter, depending on the command) to the sensed effect, a single light sensor could serve several lamps in the entire room. If the sensor doesn't confirm the command, the sensor could be read again a few times to minimize the risk that the sensor, not the controller/command, is the feedback step that's failing. That means a probably <$20 X10 light sensor could give, say, 6 $5 lamp modules in a room at $50 something like the reliability of six Insteon lamp modules at $150. Even if the whole X10 room is also controlled by a single Insteon to improve the room's feedback, that's something like $75 instead of $150-175. For each room.
And the light sensor can be used for other features when it's not processing command feedback, like daylight detection. Or just offer a way to calibrate scenarios to actual light conditions, instead of blindly switching lights on/off, so the scenario can say "make the room
X bright, using these lamps A, B, D, E, F in this proportion 1:2:1:1:2" (if dimmers are available), or other complex subjective
effects that really control the room (and the mood).
All of which just depends on the logic being included in the automation SW that adds actual intelligence to the automation. If X10's limited reliability is still enough to support the feedback loop, then it could be both cheaper (or more precisely deployed, to more individual devices) and more effective than Insteon.
OTOH, even if it's not, I still like the idea of setting room brightness by actual sensed level, not just the guess of how bright the lights would make the room if several were turned on, which unevenly lights the room to a guessed average brightness. Even if X10 is too crude to make it work, the higher expense of Insteon support of control in those terms would probably still be worth the big bucks. And feedback is the way to control most devices, like say a heated pool (or bathtub) which shouldn't just be switched on, but rather heated until it reaches a sensed temperature (or a filled level). A general facility for automation that really makes it easy to use and a pleasure to live in.