I seem to be able to replicate this behaviour.
Of course, you could always rewrite the function along the following lines (for your example):
nulldata 100
genr scalar lb=-5
genr scalar ub=1
genr series Z1 = randgen(i,0,ub-lb)+lb

HTH,

Jan Annaert

Date: woensdag 6 augustus 2014 05:31
To: "gretl-users@lists.wfu.edu" <gretl-users@lists.wfu.edu>
Subject: Re: [Gretl-users] Generate Discrete Uniform with negative bounds

Good afternoon. I tried to generate discrete uniform random variables
using randgen, but it appears that the function cannot take negative
values as bounds for the distribution (if both minimum and maximum are
positive, then everything works fine. If either one or both bounds are
negative numbers, it goes berserk).
The script is

nulldata 100
genr scalar lb=-5
genr scalar ub=1
genr series Z1 = randgen(i,lb,ub)

and the summary statistics are

Summary statistics, using the observations 1 - 100
for the variable 'Z1' (100 valid observations)

Mean                    2.7488e+009
Median                  4.2950e+009
Minimum                     0.00000
Maximum                 4.2950e+009
Standard deviation      2.0720e+009
C.V.                        0.75378
Skewness                   -0.58333
Ex. kurtosis                -1.6597
5% percentile               0.00000
95% percentile          4.2950e+009
Interquartile range     4.2950e+009
Missing obs.                      0

--