How does xpath work in lxml?

Question

Python

Hello, there was a misunderstanding how xpath works in lxml.

import lxml.html as parser

text = '''
<div id="main">
  <div>
    <a href="#">1</a>
  </div>
  
  <div>
    <a href="#">2</a>
  </div>
  
  <div>
    <a href="#">3</a>
  </div>
  
  <div>
    <a href="#">4</a>
  </div>

</div>
'''

def get_info():
    html = parser.document_fromstring(text)

    divs = html.xpath('//div[@id="main"]/div')

    print(len(divs))#4

    div = divs[0].xpath('///a')

    print(len(div))#4

if __name__ == '__main__':
    get_info()

The first xpath outputs everything correctly. There are 4 divs in the html code.
Why then, when I search using xpath in a div, does xpath search again in the entire html code, and not relative to the given div? How to do it right?