Weird features mismatch using SURF

Hi everyone.

I'm trying to use SURF to detect an object (particularly a cur number plate) in another image. Everything works perfect on MATLAB's official example but breaks on my own images.

The original images are in JPEG format and were taken on smartphone. Since tutorial's images were also in JPEG i don't think that problem is with the format.

I tried all feature detection methods that i could find in MATLAB: SURF, FAST, Harris, MSER, MinEigen and BRISK. They all worked terribly bad. Since my question is about SURF here's an example of what's going on. In my opinion SURF was able to find enough feature points on both images. Moreover it seems to have found the same feature points in same places. Yet when matchFeatures is called then no matches are found. Can anyone give any hint what could the problem be?

I have two ideas: something wrong happens when i convert my JPEG to grayscale (i use rgb2gray as offered by MATLAB console messages) or perhaps there is some specifics with how phone cameras' work. The latter might sound a bit stupid but i've also tried OpenCV and everything worked nice with webcam (SURF was able to detect a bottle label even when half of the label was not seen and all the work was done in real time) but again when phone images were fed to the code it worked badly. And i used three different phones: all results were bad both with MATLAB and OpenCV.

Thanks in advance for any ideas.