Measured energy differs from stated energy
Laboratory analyses of frequently ordered restaurant meals have reported energy contents that exceed stated values, with the largest absolute differences in the largest meals. Portion variability between locations of the same restaurant was itself a substantial source of the spread.
Where no figure is published at all, which covers most independent restaurants, there is nothing to compare against in the first place.
People underestimate, and underestimate more as meals get larger
Consumer studies have reported systematic underestimation of restaurant meal energy, and that the size of the underestimate grows with the size of the meal. The intuition scales worse than the food does.
What is knowable and what is not
For a restaurant plate, the composition is visible and the quantity is not. You can see there is rice, chicken, a sauce and oil. You cannot see 180 g versus 260 g, and neither can any tool working from a photograph alone.
That is the honest boundary. A photo captures what a typed description leaves out: the sauce, the side, the oil. It remains weakest exactly where portion size decides the total.
The practical consequence
Logging a restaurant meal is an estimate no matter how it is done. What is worth optimising is not the illusion of precision on that one meal, but whether it gets logged at all. A skipped entry is a 100 % error, and it is the most common one.