The short answer is no! You cannot extrapolate from close proximity E field (or B field) measurements to a proper test set up in a lab. Near field is the region where em radiation need not be both E and B fields, but in the far field, after some propagation distance proportional to the wave length, the E and B fields merge (as in they are both present). You can have a radiator emmiting B fields which are invisible to an E field probe up close, but would be picked up by an E field probe in the far field measurement. The reverse is also possible with B field probes.
The long answer is it depends on the source, the fix, and the testing standard.
You need power and antennas to radiate at any given frequency, which means in practice you can either have high currents or long current paths causing your failure. Given the high frequency I'm going to guess a good radiator rather than a high current. If your fix involved snubbing or terminating (ie, stopping the noise at source) then you've got a good chance of seing some improvement, but if your fix involved tin foil or ferrites (ie, blocking the noise from getting to your probe) i would say its 50/50.
Be aware that the test houses I've worked with have a 6dB error tolerance, and things can easily change by that much just by a different setup of the same equipment. (Different cable routing etc). 3dB isn't all that much in the grand scheme of things. It could easily be just the probe on your bench in a slightly different position.