Problem with ncgen and/or ncdump

Hello,

I've run into a problem with ncgen/ncdump that perhaps someone can help me
with.

I'm running ncgen to create a netCDF file:

    ncgen -b -o somefile someinput

The "someinput" file contains text like this:

         float liq(time) ; 
                liq:long_name = "Total liquid water along LOS path" ; 
                liq:units = "cm" ; 
                liq:resolution = 0.0001f ; 
                liq:missing_value = -9999.f ; 

Ncgen works just fine, and when I run

        ncdump somefile

I see:

        float liq(time) ;
                liq:long_name = "Total liquid water along LOS path" ;
                liq:units = "cm" ;
                liq:resolution = 9.9999997e-05f ;
                liq:missing_value = -9999.f ;
                liq:uncertainty = "+/- 0.002 cm" ;

Note that the "resolution" attribute has been changed. It has essentially
the same value as the "someinput" information, but it is not as easy to
read.

My questions:

1) Is there any way to "force" ncgen and/or ncdump to display the attribute
info as I want it to?
2) Is there a better way to define the "resolution" attribute in the first
place?

Any help will be appreciated. Interestingly enough, this does not always
happen.  (This corruption -- if you can call it that -- of the resolution
attribute on one system that I have access to.)