Saxon

The XPath 2.0 specification here is incompatible with XPath 1.0; XPath 2.0 changed the rules so that xs:double values outside the range 1e-6 to 1e+6 will always be output in exponential form. It's easy to get numbers in integer form simply by casting to xs:integer before converting to a string, or by using format-number to control the formatting more precisely. Ideally, avoid using xs:double in the first place: for example, if @page has the lexical form of an integer, use xs:integer(@page)+1 rather than @page+1.

One thing I did notice in your code is very extensive use of constructs like

                <xsl:with-param name="PadNumber">
                  <xsl:value-of select="$SAMPLESHT1FRONTStartNbr1" />
                </xsl:with-param>
                <xsl:with-param name="PadString">
                  <xsl:value-of select="'0000000'" />
                </xsl:with-param>
                <xsl:with-param name="FieldLength">
                  <xsl:value-of select="7" />
                </xsl:with-param>

It's always better to write this as

                <xsl:with-param name="PadNumber" select="$SAMPLESHT1FRONTStartNbr1" />
                <xsl:with-param name="PadString" select="'0000000'" />
                <xsl:with-param name="FieldLength" select="7" />

not just for brevity, but because you really don't want these variables to be result tree fragments. Saxon tries hard to optimize the more verbose form, but the two forms aren't 100% equivalent in edge cases, so there's a limit to what's possible. Ideally, add a type declaration (e.g. as="xs:integer") as well, which gives you better diagnostics, better documentation, and better performance.