Unless your sample is extremely strong you could set a fixed RG and leave it. Alternatively, you could perform an RGA but add a line in your AU program to put a maximum value. In my studies, a receiver gain of more than 128 does almost nothing for signal to noise and can cause peak distortion problems (and bad baselines). Try:

float myrg;
GETCURDATA
RGA
FETCHPARS("RG",&myrg)
if (myrg>128) myrg=128;
STOREPAR("RG",myrg)
QUIT

From my experience, if the S/N is poor, or if the lines are very broad, RGA will give nonsensical values. This may or may not be your problem as my experience is mainly with solid-state NMR machines with samples that have much broader linewidths.

For checking whether the RG is set too high, look at the FID as the start of the experiment as it is acquiring and see if the top and bottom looks horizontal (cut off), if it is then it is too high and will give you distortions, as shown on this page written by Glenn Facey of University of Ottawa: http://u-of-o-nmr-facility.blogspot.ca/2007/10/consequences-of-setting-receiver-gain.html Unfortunately you can only see this at the beginning of the experiment as once you have acquired a few more scans the distortions in the FID look as if they are smoothed out but will still give you a distorted spectrum.

The solution, as apophis said, is to set it manually. This page (also by Glenn Facey) explains how S/N can vary with RG. You should vary the value of RG manually, trying several different values, and find a maximum that gives good S/N without distortions. http://u-of-o-nmr-facility.blogspot.co.uk/2013/02/receiver-gain-and-signal-to-noise-ratio.html

Hey!

I also had this problem many times. Unfortunately i can not give u a good explanation for this problem but it is bruker, very shitty stuff. Just decrease the rg manually to a proper val (~500) your spec may look better. Hope this may help u.