How preceding-sibling works in XPath and Python? It seems to display wrong output

For the XML data

<X>
 <Y1>ABC1</Y1>
 <Y2>ABC2</Y2>
 <Z>
   <T>
     <R1>ABC3</R1>
     <R2>ABC4</R2>
   </T>
   <T>
     <R1>ABC5</R1>
     <R2>ABC6</R2>
   </T>
 </Z>
 <Y1>ABC7</Y1>
 <Y2>ABC8</Y2>
 <Z>
   <T>
     <R1>ABC3</R1>
     <R2>ABC9</R2>
   </T>
   <T>
     <R1>ABC5</R1>
     <R2>ABC9</R2>
   </T>
 </Z>
</X>

I wrote a sample python file like the below.

from lxml import etree
tree = etree.parse('test.xml')
for i in tree.xpath("//X/Z/T[R1='ABC3']/parent::*/preceding-sibling::*"):
    print(i.tag, " - ", i.text)

I expected output like

Y1  -  ABC1
Y2  -  ABC2
Y1  -  ABC1
Y2  -  ABC2
Z  -  

Y1  -  ABC7
Y2  -  ABC8

but received one like

Y1  -  ABC1
Y2  -  ABC2
Z  -  

Y1  -  ABC7
Y2  -  ABC8

It should print all preceding sibling. For 1st match of "R1=ABC3",it should print Y1 and Y2. For 2nd match of "R1=ABC", it should print the 5 siblings. Total 7 elements should be printed.
What is the error here?

XPath 1.0 has a concept of node-sets where each / step eliminates duplicates based on node identity so a single XPath expression as you have used will not give a set that contains the same node twice, any duplicates are eliminated.

In XPath 2.0, while of course the / step operator continues to have the same duplicate elimination semantics, there is a more generalized concept of sequences using for .. return (for $p in //X/Z/T[R1='ABC3']/parent::* return $p/preceding-sibling::*) or in XPath 3.1 ! (//X/Z/T[R1='ABC3']/parent::*!preceding-sibling::*) that would allow you to include duplicates, see https://xqueryfiddle.liberty-development.net/eiZQFoV.

In XPath 1.0 you would need to use several XPath evaluations in a loop of the host language (e.g. Python) or in the case of Python you could use list comprehensions element_list = [el for parent in tree.xpath("//X/Z/T[R1='ABC3']/parent::*") for el in parent.xpath("preceding-sibling::*")]
.

The question is tagged xslt, but you're not using XSLT. The expected output can be achieved using the following stylesheet:

XSLT 1.0

<xsl:stylesheet version="1.0" 
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output method="text" encoding="utf-8"/>

<xsl:template match="/X">
	<xsl:for-each select="Z[T/R1='ABC3']">
		<xsl:for-each select="preceding-sibling::*">
			<xsl:value-of select="name()" />
			<xsl:text> - </xsl:text>
			<xsl:value-of select="text()" />
			<xsl:text>
</xsl:text>
		</xsl:for-each>
	</xsl:for-each>
</xsl:template>

</xsl:stylesheet>

As noted in the answer by Martin Honnen, it is necessary to process the preceding siblings of each matched node separately, in order to get two separate lists.

Note also that your expression:

Z/T[R1='ABC3']/parent::* 

is unnecessarily convoluted: clearly, the parent of the matched T must be Z - so you can write simply:

Z[T/R1='ABC3']