diff --git a/lib/cartodb/models/aggregation/aggregation-query.js b/lib/cartodb/models/aggregation/aggregation-query.js
index 23028f1b..4c2c34bf 100644
--- a/lib/cartodb/models/aggregation/aggregation-query.js
+++ b/lib/cartodb/models/aggregation/aggregation-query.js
@@ -248,28 +248,58 @@ const gridResolution = ctx => {
     return `GREATEST(${256/ctx.res}*CDB_XYZ_Resolution(CDB_ZoomFromScale(!scale_denominator!)), ${minimumResolution})::double precision`;
 };
 
+// Each aggregation cell is defined by the cell coordinates Floor(x/res), Floor(y/res),
+// i.e. they include the West and South borders but not the East and North ones.
+// So, to avoid picking points that don't belong to cells in the tile, given the tile
+// limits Xmin, Ymin, Xmax, Ymax (bbox), we should select points that satisfy
+// Xmin <= x < Xmax and Ymin <= y < Ymax (with x, y from the_geom_webmercator)
+// On the other hand we can efficiently filter spatially (relying on spatial indexing)
+// with `the_geom_webmercator && bbox` which is equivalent to
+//   Xmin <= x <= Xmax and Ymin <= y <= Ymax
+// So, in order to be both efficient and accurate we will need to use both
+// conditions for spatial filtering.
+const spatialFilter = `
+  (_cdb_query.the_geom_webmercator && _cdb_params.bbox) AND
+    ST_X(_cdb_query.the_geom_webmercator) >= _cdb_params.xmin AND ST_X(_cdb_query.the_geom_webmercator) < _cdb_params.xmax AND
+    ST_Y(_cdb_query.the_geom_webmercator) >= _cdb_params.ymin AND ST_Y(_cdb_query.the_geom_webmercator) < _cdb_params.ymax
+`;
+
 // Notes:
 // * We need to filter spatially using !bbox! to make the queries efficient because
 //   the filter added by Mapnik (wrapping the query)
 //   is only applied after the aggregation.
 // * This queries are used for rendering and the_geom is omitted in the results for better performance
+// * If the MVT extent or tile buffer was 0 or a multiple of the resolution we could use directly
+//   the bbox for them, but in general we need to find the nearest cell limits inside the bbox.
+const sqlParams = (ctx) => `
+_cdb_res AS (
+    SELECT
+    ${gridResolution(ctx)} AS res,
+    !bbox! AS bbox
+),
+_cdb_params AS (
+    SELECT
+      res,
+      bbox,
+      CEIL(ST_XMIN(bbox)/res)*res AS xmin,
+      FLOOR(ST_XMAX(bbox)/res)*res AS xmax,
+      CEIL(ST_YMIN(bbox)/res)*res AS ymin,
+      FLOOR(ST_YMAX(bbox)/res)*res AS ymax
+      FROM _cdb_res
+)
+`;
 
 // The special default aggregation includes all the columns of a sample row per grid cell and
 // the count (_cdb_feature_count) of the aggregated rows.
 const defaultAggregationQueryTemplate = ctx => `
-    WITH
-    _cdb_params AS (
-        SELECT
-        ${gridResolution(ctx)} AS res,
-        !bbox! AS bbox
-    ),
+    WITH ${sqlParams(ctx)},
     _cdb_clusters AS (
         SELECT
             MIN(cartodb_id) AS cartodb_id
             ${dimensionDefs(ctx)}
             ${aggregateColumnDefs(ctx)}
         FROM (${ctx.sourceQuery}) _cdb_query, _cdb_params
-        WHERE _cdb_query.the_geom_webmercator && _cdb_params.bbox
+        WHERE ${spatialFilter}
         GROUP BY
             Floor(ST_X(_cdb_query.the_geom_webmercator)/_cdb_params.res),
             Floor(ST_Y(_cdb_query.the_geom_webmercator)/_cdb_params.res)
@@ -284,12 +314,7 @@ const defaultAggregationQueryTemplate = ctx => `
 
 const aggregationQueryTemplates = {
     'centroid': ctx => `
-        WITH
-        _cdb_params AS (
-            SELECT
-            ${gridResolution(ctx)} AS res,
-            !bbox! AS bbox
-        )
+        WITH ${sqlParams(ctx)}
         SELECT
             MIN(_cdb_query.cartodb_id)  AS cartodb_id,
             ST_SetSRID(
@@ -301,7 +326,7 @@ const aggregationQueryTemplates = {
             ${dimensionDefs(ctx)}
             ${aggregateColumnDefs(ctx)}
         FROM (${ctx.sourceQuery}) _cdb_query, _cdb_params
-        WHERE _cdb_query.the_geom_webmercator && _cdb_params.bbox
+        WHERE ${spatialFilter}
         GROUP BY
             Floor(ST_X(_cdb_query.the_geom_webmercator)/_cdb_params.res),
             Floor(ST_Y(_cdb_query.the_geom_webmercator)/_cdb_params.res)
@@ -310,12 +335,7 @@ const aggregationQueryTemplates = {
     `,
 
     'point-grid': ctx => `
-        WITH
-        _cdb_params AS (
-            SELECT
-            ${gridResolution(ctx)} AS res,
-            !bbox! AS bbox
-        ),
+        WITH ${sqlParams(ctx)},
         _cdb_clusters AS (
             SELECT
             MIN(_cdb_query.cartodb_id) AS cartodb_id,
@@ -324,7 +344,7 @@ const aggregationQueryTemplates = {
             ${dimensionDefs(ctx)}
             ${aggregateColumnDefs(ctx)}
             FROM (${ctx.sourceQuery}) _cdb_query, _cdb_params
-            WHERE the_geom_webmercator && _cdb_params.bbox
+            WHERE ${spatialFilter}
             GROUP BY _cdb_gx, _cdb_gy ${dimensionNames(ctx)}
             ${havingClause(ctx)}
         )
@@ -337,19 +357,14 @@ const aggregationQueryTemplates = {
     `,
 
     'point-sample': ctx => `
-        WITH
-        _cdb_params AS (
-            SELECT
-            ${gridResolution(ctx)} AS res,
-            !bbox! AS bbox
-        ),
+        WITH ${sqlParams(ctx)},
         _cdb_clusters AS (
             SELECT
                 MIN(cartodb_id) AS cartodb_id
                 ${dimensionDefs(ctx)}
                 ${aggregateColumnDefs(ctx)}
             FROM (${ctx.sourceQuery}) _cdb_query, _cdb_params
-            WHERE _cdb_query.the_geom_webmercator && _cdb_params.bbox
+            WHERE ${spatialFilter}
             GROUP BY
                 Floor(ST_X(_cdb_query.the_geom_webmercator)/_cdb_params.res),
                 Floor(ST_Y(_cdb_query.the_geom_webmercator)/_cdb_params.res)