When the absorbance is that high, almost no light is being transmitted through your sample. Since absorbance = log(I0/It), where I0 and It are the incident and transmitted intensities respectively, small random fluctuations in the transmitted signal (due to e.g. stray light or instrumental noise) will give rise to large variations in the apparent absorbance. To measure absorbance accurately in this region you will need to dilute your sample - I suggest by at least a factor of 100, if you are getting an absorbance of 4. (And if you're using a 10 mm cell, try using a 1 mm cell.)