Counts are exact, but the three groups are nowhere near the same size — 63 men, 45 women and 5 others overall. A male bar will run longer for that reason alone. Switch to "% of each group" whenever you are judging whether men and women differ; use counts when you want to know how many people are actually behind a bar.
The grey bar is five people — two non-binary, one transgender, and two who preferred not to say. They are shown so the chart accounts for everyone who answered, but five people cannot support a percentage: a single response moves a grey bar by 20 points. Read the grey counts, never the grey shape, and never quote a grey percentage on its own.
Multi-select questions don't sum to the group size. Dealbreakers, lifestyle, compatibility, reasons to ignore a post, marrying across, and the religion questions all allowed several ticks per person.
The religion questions branched , so only respondents raised in that faith saw each one: 88 Hindu, 15 Muslim, 13 Christian, 8 Sikh, 6 Jain. Split by sex and filtered by age, several of those cells fall to one or two people. They are shown for completeness, not as findings.
Cleaning applied. "IT" merged into "Software / IT / Tech"; lowercase "unemployed" merged into "Unemployed"; "others" relabelled "Other / unspecified". Free-text write-ins are excluded — none exceeded two people. Religion option labels contain commas and are matched as whole strings rather than split.