Variable yields in T7 trascritpion

[the End]

I am finding enormous variability in terms of yield of specific labeled RNA's
from day to day in basic T7 transcriptions. I have compared various 
buffers (80 mM Hepes, 40 mM TrisOac, 40 mM Tris-Cl) and juggled other
components (BSA, NP-40, spermidine, Pyrophosphatase, RNAsin) and reaction 
times (1h, 1.5h, 2 h) as well as nucleotide levels appropriate for 
making internally labeled transcripts (50-200 uM limiting nucleotide
and 0.5-3 mM of the other three).

Overall, a particular working setup and template make work well one time
in four or five.

I believe that differnces may be related to low nucleotide levels, as 
sythesizing cold RNA shows variability, but far less completely dismal
yields.

I have also compared PCR and plasmid templates, with various additional 
bases upstream of the T7 promoter.

Anyone seeing this kind of hit or miss synthesis ? I use NEB T7 pol
which may not be as concentrated as that found in the various kits.
Epicentre's "secret" buffer-enzyme combo has received rave reviews
across the hall, but apparently they most commonly make unlabeled 
RNA's at high NTP levels.

Any thing else I haven't tried ?



Thanks,