The energy from the light source will fall off as 1/r2 (for a source that's relatively small). Beyond this, if everything else is held constant, the only problem could be non-linearity in the sensor.
To check this with your data, you would expect E x r2=const, and this roughly holds for your data:
110 x 52 = 2750, and
70 x 62 = 2520,
so these are within 10% which seems fairly close, so it looks like the basic rule will hold.
Non-linear sensors are common, so you should be sure to check this over the full range that you'll be using it. But if it's a linear sensor, the other issues that people are mentioning (e.g. reflective surfaces) won't be a problem because, for light transmission and reflection, everything (almost) is linear and will therefore be intrinsically compensated for by a single calibration constant. The angle of the light source, absorbing materials, etc, all won't matter as long as they don't change.
If you test a few points, including the extremes of the range you're interested in, and it follows the 1/r2 rule, you're good to go. Then, of course, calculate what the const is, and r = sqrt(const/E).