500 kHz (watch the capitalization: Khz is wrong!) is a very low sample rate for the N210. In fact, the N210 has to decimate by a factor of 200 from its ADC's sampling rate, and that might introduce unexpected roll-off and aliasing, because that over-stresses the CIC decimator (if this doesn't make too much sense to you: don't worry, it's an implementation detail. Just: rather go for 1 or 2 MS/s sampling rate and do the rest of the decimation in software).
Also make sure you're actually seeing the signal you think you're seeing: your USRP AND your LoRa transmitter might both have frequency offsets. But: 15 MHz difference at 915 MHz? That would be an error of 1/61 ; the USRP has a frequency accuracy of ca 10 ppm (that's 1/100000), your LoRa transmitter maybe of 50 ppm. No way you're observing a 915 MHz signal at 930 MHz. Either this is something else, or you're misinterpreting something and your transmitter isn't trying to transmit at 915 MHz!
It's hard to tell from your slightly unusual choice of color scheme whether your spectral broadening happens from an overdrive of the receive amplifier. Reduce gain – if it gets better from that, that's happening.
Also make sure you're not actually seeing what you want to see – after all, LoRa is chirp-spread spectrum, and you're sampling these chirps at more or less irregular intervals. How does your spectrum change when you change the FFT size?
