When comparing the pixel colours from glClear and normal rendering it was found 
that the same colour differed by one pixel value. This was found to be caused 
by glClear (in s_buffers.c, function clear_rgba_buffer) using the macro 
FLOAT_TO_USHORT which truncates while the normal rendering pipeline (in 
t_vertex_generic.c) uses the macro UNCLAMPED_FLOAT_TO_USHORT which rounds. So 
for a pixel value expected to be 2.52 the glClear calculated 2 and the 
rendering pipeline calculated 3.

This was resolved simply resolved by substituting the latted macro in the 
glClear case. Therefore the lines 135 to 138 were changed to 

         UNCLAMPED_FLOAT_TO_USHORT(clear16[0], ctx->Color.ClearColor[0]);
         UNCLAMPED_FLOAT_TO_USHORT(clear16[1], ctx->Color.ClearColor[1]);
         UNCLAMPED_FLOAT_TO_USHORT(clear16[2], ctx->Color.ClearColor[2]);
         UNCLAMPED_FLOAT_TO_USHORT(clear16[3], ctx->Color.ClearColor[3]);

This resolved the problem for this specific instance.